gh-208: new configuration format

[ntessore]

Implements a new "configuration" format. This can be used for both CLI and pipeline. Still work in progress.

Closes: #208

[ntessore]

After discussion with @JaimeRZP today: here is the proposed config in TOML form:

# example config file for Heracles

[[catalogs]]
fields = ["POS", "VIS"]
label = "position catalogue"
selections = [
    {key = 1, selection = "POS_TOM_BIN_ID==1", visibility = "visibility.1.fits"},
    {key = 2, selection = "POS_TOM_BIN_ID==2", visibility = "visibility.2.fits"},
    {key = 3, selection = "POS_TOM_BIN_ID==3", visibility = "visibility.3.fits"},
]
source = "pos_catalog.fits"

[[catalogs]]
fields = ["SHE", "WHT"]
label = "shear catalogue"
selections = [
    {key = 1, selection = "TOM_BIN_ID==1"},
    {key = 2, selection = "TOM_BIN_ID==2"},
    {key = 3, selection = "TOM_BIN_ID==3"},
]
source = "she_catalog.fits"
visibility = "coverage.fits" # global visibility for this catalogue

[[fields]]
key = "POS"
type = "positions"
columns = ["RIGHT_ASCENSION", "DECLINATION"]
mask = "VIS"
mapper = "healpix" # default, requires healpy
lmax = 2000
nside = 2048

[[fields]]
key = "VIS"
type = "visibility"
lmax = 6000 # maximum lmax + l2max from spectra involving POS
nside = 4096

[[fields]]
key = "SHE"
type = "shears"
columns = ["SHE_RA", "SHE_DEC", "SHE_E1_CAL", "-SHE_E2_CAL", "SHE_WEIGHT"]
mask = "WHT"
mapper = "discrete" # requires ducc, no nside here
lmax = 3000

[[fields]]
key = "WHT"
type = "weights"
columns = ["SHE_RA", "SHE_DEC", "SHE_WEIGHT"]
lmax = 8000 # maximum lmax + l2max from spectra involving SHE
nside = 8192

[[spectra]]
# galaxy clustering
key = "POS-POS" # can use string syntax here
lmin = 10 # used only when doing angular binning
lmax = 2000
l2max = 4000 # mixing matrix: number of columns
l3max = 6000 # lmax + l2max is the default
bins = 32 # angular binning
spacing = "log" # angular binning
weights = "2l+1" # angular binning

[[spectra]]
# cosmic shear
key = "SHE-SHE"
lmin = 10
lmax = 3000
l2max = 5000
bins = 32
spacing = "log"
weights = "2l+1"

[[spectra]]
# galaxy-galaxy lensing
key = ["POS", "SHE"] # can also use sequence
lmin = 10
lmax = 1000
l2max = 2000
bins = 32
spacing = "log"
weights = "2l+1"

And here the exact same in YAML form:

catalogs:
- fields: [POS, VIS]
  label: position catalogue
  selections:
  - key: 1
    selection: POS_TOM_BIN_ID==1
    visibility: visibility.1.fits
  - key: 2
    selection: POS_TOM_BIN_ID==2
    visibility: visibility.2.fits
  - key: 3
    selection: POS_TOM_BIN_ID==3
    visibility: visibility.3.fits
  source: pos_catalog.fits
- fields: [SHE, WHT]
  label: shear catalogue
  selections:
  - key: 1
    selection: TOM_BIN_ID==1
  - key: 2
    selection: TOM_BIN_ID==2
  - key: 3
    selection: TOM_BIN_ID==3
  source: she_catalog.fits
  visibility: coverage.fits
fields:
- key: POS
  type: positions
  columns: [RIGHT_ASCENSION, DECLINATION]
  mask: VIS
  mapper: healpix
  lmax: 2000
  nside: 2048
- key: VIS
  type: visibility
  lmax: 6000
  nside: 4096
- key: SHE
  type: shears
  columns: [SHE_RA, SHE_DEC, SHE_E1_CAL, -SHE_E2_CAL, SHE_WEIGHT]
  mask: WHT
  mapper: discrete
  lmax: 3000
- key: WHT
  type: weights
  columns: [SHE_RA, SHE_DEC, SHE_WEIGHT]
  lmax: 8000
  nside: 8192
spectra:
- key: POS-POS
  lmin: 10
  lmax: 2000
  l2max: 4000
  l3max: 6000
  bins: 32
  spacing: log
  weights: 2l+1
- key: SHE-SHE
  lmin: 10
  lmax: 3000
  l2max: 5000
  bins: 32
  spacing: log
  weights: 2l+1
- key: [POS, SHE]
  lmin: 10
  lmax: 1000
  l2max: 2000
  bins: 32
  spacing: log
  weights: 2l+1

I find the TOML is slightly more readable, because the splitting into "groups" helps information overload. But the YAML could maybe be separated visually with additional whitespace, and I have no strong feelings one way or the other.