Refactor CI tests for speed

[janfb]

The "fast" tests we run for CI became quite slow: pytest -m "not slow and not gpu"

To speed them up I removed a couple of duplicate tests and used better MCMC parameters.
The biggest speed up comes from using slice_np_vectorized with many chains and few warm up steps (when posterior accuracy is not tested).

The speed up is about 30%.
@michaeldeistler can you confirm by testing on main and on this branch locally?

I separated commits per test file to be able to correct potential misses. We can squash the commits when merging.

TODO:

• check test coverage

[michaeldeistler]

Thanks! Running the tests now.

[michaeldeistler]

On my MBP with M3 the timings are 422.69s vs 170.27s. Awesome!

[codecov]

Codecov Report

Attention: 2 lines in your changes are missing coverage. Please review.

Comparison is base (9e224da) 75.46% compared to head (bd4c83b) 75.28%.

Files	Patch %	Lines
sbi/inference/posteriors/mcmc_posterior.py	33.33%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #923      +/-   ##
==========================================
- Coverage   75.46%   75.28%   -0.18%     
==========================================
  Files          80       80              
  Lines        6284     6284              
==========================================
- Hits         4742     4731      -11     
- Misses       1542     1553      +11     

Flag	Coverage Δ
unittests	75.28% <33.33%> (-0.18%)	⬇️

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

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.