Update demographics.py and PolicyEngine

[rickecon]

This PR is mainly to update demographics.py to use UN data. However, I also added a bunch of other related updates. This PR does the following:

• Updated demographics.py to use the UN data portal to get UK demographic data.
• Updated calibrate.py to allow for demographics.
• Added four .csv files to oguk/data/demographic/. This files allow for demographics.py to have the option to not download the data from the UN data portal.
• Updated get_micro_data.py and test_get_micro_data.py with PolicyEngine references instead of OpenFisca.
• Updated environment.yml and setup.py.
• Updated run_oguk.py with more consistent specification, and updated references from openfisca to policyengine.
• Small updates to .gitignore, README.md, demographics.rst, get_micro_data.rst, and tax_functions.md.
• Deleted pyproject.toml which was just a reference to the black package.
• Updates the CHANGELOG.md and updates the version number in setup.py.
• Updates the GitHub Actions deploy_docs.yml, docs_check.yml, and build_and_test.yml.

cc: @jdebacker

[rickecon]

@jdebacker @nikhilwoodruff. The demographics.py module works great. I have tested all of its functions. However, I have not been able to successfully run the run_oguk.py example script. The error I am getting seems to be related to the import statement from policyengine_core.reforms import Reform. Here is the Traceback.

Traceback (most recent call last):
  File "/Users/richardevans/Docs/Economics/OSE/OG-UK/./examples/run_oguk.py", line 7, in <module>
    from policyengine_core.reforms import Reform
  File "/Users/richardevans/opt/anaconda3/envs/oguk-dev/lib/python3.9/site-packages/policyengine_core/__init__.py", line 1, in <module>
    from policyengine_core.simulations import Microsimulation, Simulation
  File "/Users/richardevans/opt/anaconda3/envs/oguk-dev/lib/python3.9/site-packages/policyengine_core/simulations/__init__.py", line 9, in <module>
    from .microsimulation import Microsimulation
  File "/Users/richardevans/opt/anaconda3/envs/oguk-dev/lib/python3.9/site-packages/policyengine_core/simulations/microsimulation.py", line 9, in <module>
    from policyengine_core.simulations.simulation import Simulation
  File "/Users/richardevans/opt/anaconda3/envs/oguk-dev/lib/python3.9/site-packages/policyengine_core/simulations/simulation.py", line 9, in <module>
    from policyengine_core import commons, periods
  File "/Users/richardevans/opt/anaconda3/envs/oguk-dev/lib/python3.9/site-packages/policyengine_core/commons/__init__.py", line 1, in <module>
    from .formulas import apply_thresholds, concat, switch
  File "/Users/richardevans/opt/anaconda3/envs/oguk-dev/lib/python3.9/site-packages/policyengine_core/commons/formulas.py", line 13, in <module>
    from policyengine_core.parameters.parameter_node import ParameterNode
  File "/Users/richardevans/opt/anaconda3/envs/oguk-dev/lib/python3.9/site-packages/policyengine_core/parameters/__init__.py", line 15, in <module>
    from .operations import (
  File "/Users/richardevans/opt/anaconda3/envs/oguk-dev/lib/python3.9/site-packages/policyengine_core/parameters/operations/__init__.py", line 1, in <module>
    from .get_parameter import get_parameter
  File "/Users/richardevans/opt/anaconda3/envs/oguk-dev/lib/python3.9/site-packages/policyengine_core/parameters/operations/get_parameter.py", line 2, in <module>
    from policyengine_core.parameters.parameter_node import ParameterNode
  File "/Users/richardevans/opt/anaconda3/envs/oguk-dev/lib/python3.9/site-packages/policyengine_core/parameters/parameter_node.py", line 6, in <module>
    from policyengine_core import commons, parameters, tools
  File "/Users/richardevans/opt/anaconda3/envs/oguk-dev/lib/python3.9/site-packages/policyengine_core/tools/__init__.py", line 6, in <module>
    import numexpr
  File "/Users/richardevans/opt/anaconda3/envs/oguk-dev/lib/python3.9/site-packages/numexpr/__init__.py", line 24, in <module>
    from __config__ import show as show_config, get_info
ModuleNotFoundError: No module named '__config__'

[nikhilwoodruff]

OK @rickecon - think this is getting somewhere: the package installs are now working, and the microdata tests passing, but the run_oguk module is hitting an error, and I don't really understand it but it seems plausibly logic-related:

Warning: There is a discrepency between the start year of the model and that of the tax functions!!p.BW = 6, BW_in_tax_params = 80
Warning: There is a discrepency between the ages used in the model and those in the tax functions!!p.S = 80, S_in_tax_params = 6

So I think the BW in p is correct, right? Is the tax function estimation putting things in the wrong place? Thought that was unlikely because it's a call to OG-Core, but yeah any ideas?

Also, here's the shape of each item in the tax function parameter dictionary:

{
  'tfunc_etr_params_S': (6, 80, 1), 
  'tfunc_mtrx_params_S': (6, 80, 1), 
  'tfunc_mtry_params_S': (6, 80, 1), 
  'tfunc_avginc': (6,), 
  'tfunc_avg_etr': (6,), 
  'tfunc_avg_mtrx': (6,),
  'tfunc_avg_mtry': (6,), 
  'tfunc_frac_tax_payroll': (6,), 
  'tfunc_etr_sumsq': (6, 80), 
  'tfunc_mtrx_sumsq': (6, 80), 
  'tfunc_mtry_sumsq': (6, 80), 
  'tfunc_etr_obs': (6, 80),
  'tfunc_mtrx_obs': (6, 80), 
  'tfunc_mtry_obs': (6, 80), 
  'tfunc_time': (), 
  'tax_func_type': (), 
  'start_year': (), '
  BW': ()
}

[codecov]

Codecov Report

Base: 70.06% // Head: 72.91% // Increases project coverage by +2.84% 🎉

Coverage data is based on head (8b72361) compared to base (a72a51e).
Patch coverage: 59.09% of modified lines in pull request are covered.

❗ Current head 8b72361 differs from pull request most recent head a2d574c. Consider uploading reports for the commit a2d574c to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #57      +/-   ##
==========================================
+ Coverage   70.06%   72.91%   +2.84%     
==========================================
  Files           4        4              
  Lines         147      144       -3     
==========================================
+ Hits          103      105       +2     
+ Misses         44       39       -5     

Flag	Coverage Δ
unittests	72.91% <59.09%> (+2.84%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
setup.py	0.00% <0.00%> (ø)
oguk/get_micro_data.py	72.46% <70.58%> (+0.40%)	⬆️
oguk/tests/test_get_micro_data.py	87.80% <100.00%> (+0.30%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[nikhilwoodruff]

Just a few thoughts before our next OG call:

• The steady-state is now consistently solving under linear tax functions (see non-last action test passes)
• The steady-state is now consistently solving for me under DEP tax functions (but seems to be failing on Windows- see the last action)
• All tests apart from the Windows test and the documentation build test are passing
• The docs build test is failing because of a dependency error I've seen elsewhere (I think it's an issue with JupyterBook, but the fix I used for other repos, force-installing jsonschema[format-nongpl], didn't work here for some reason)

I'd recommend merging this PR without the Windows and docs build test passes because neither of them represent a regression really (the docs just won't get updated rather than being taken down), and also because I think this PR overlaps/aids some of the others (e.g. I think it fixes #55, and resolves pretty much all PolicyEngine-UK-related issues in other PRs).

[rickecon]

@nikhilwoodruff. I discovered that one reason my earlier runs of the model were breaking in the get_micro_data.py stage (see this output in the comment above) is that PolicyEngine-UK doesn't work with the Anaconda Python distribution for Mac OS (64-bit M1 optimized distribution). We found that this distribution sped up computation times by 33% (increase in computation going from standard Mac OS Python distribution to M1 optimized is 50%). For the time being, I have reverted my Python distribution to the standard Anaconda distribution.

However, I now get the following error when running on my machine in the oguk-dev conda environment which says that ReformType is not a defined type in line 45 of get_micro_data.py.

(oguk-dev) richardevans@Richards-MacBook-Pro OG-UK % python ./examples/run_oguk.py
WARNING:root:
    Could not locate FRS microdata. If you have access to the data, try running:

    policyengine-uk-data enhanced_frs download 2022
    
WARNING:root:Using synthetic FRS microdata.
INFO:root:Downloading 2022 synthetic FRS microdata.
100%|█████████████████████████████████████████████████████████████████| 27.0M/27.0M [00:03<00:00, 8.58MiB/s]
Traceback (most recent call last):
  File "/Users/richardevans/Docs/Economics/OSE/OG-UK/./examples/run_oguk.py", line 9, in <module>
    from oguk.calibrate import Calibration
  File "/Users/richardevans/Docs/Economics/OSE/OG-UK/oguk/calibrate.py", line 2, in <module>
    from oguk import get_micro_data
  File "/Users/richardevans/Docs/Economics/OSE/OG-UK/oguk/get_micro_data.py", line 45, in <module>
    reform: ReformType,
NameError: name 'ReformType' is not defined

[nikhilwoodruff]

@rickecon - I had no idea about the M1-optimised Python distribution, interesting! (I think we might be on the same Mac device?).

Hmm, I thought I'd removed the ReformType reference, it's no longer in the repo for me?