desi_zcatalog refactor

[sbailey]

This PR significantly updates desi_zcatalog to add parallelism and make it easier to use for productions. Relative to current main, this PR changes the script options/usage but does not change the actual outputs other than their sort order. The code changes are big enough that it might be easiest to evaluate the final state and outputs rather than the code diff.

Usage

The updated script supports two modes:

• desi_zcatalog --indir ... : recursively find redrock*.fits files under an arbitrary input directory and stack them. This is useful for creating test redshift catalogs from custom runs of redrock that might be organized outside of the context of a production.
• desi_zcatalog --group healpix|cumulative|pernight|perexp --survey SURVEY --program PROGRAM ... uses a standard specprod structure to group either by healpix or tiles.

Current main requires a pre-step to generate an ASCII list of tiles to include for each survey,program combination which is a hassle; this PR updates the script to use the tiles file from the production and filter by (survey, program) on-the-fly.

Example outputs

See /pscratch/sd/s/sjbailey/desi/dev/zcat_options/

ztile-sv2-dark-cumulative-main.fits, produced with current main with:

desi_zcatalog --add-units --recoadd-fibermap --ztile \
    --group cumulative \
    --indir /global/cfs/cdirs/desi/spectro/redux/iron/tiles/cumulative/ \
    --tiles sv2-dark-tiles.txt \
    --outfile ztile-sv2-dark-cumulative-main.fits

ztile-sv2-dark-cumulative-pr.fits, produced with this branch with:

export SPECPROD=iron
desi_zcatalog --add-units --recoadd-fibermap --nproc 16 \
    --group cumulative \
    --survey sv2 --program dark \
    --outfile ztile-sv2-dark-cumulative-pr.fits

fitsdiff verifies that these files are the same except for the CHECKSUM/DATASUM comment timestamps, and the CHECKSUM value (due to the differing timestamps leading to differing CHECKSUMs).

Detail: main stacks the tiles in whatever order they appear in the --tiles TILES option; this PR stacks them in whatever order they appear in the $DESI_SPECTRO_REDUX/$SPECPROD/tiles-$SPECPROD.fits file. In this case I made them match, but in general they could be in a different order. The compare.py script in the same directory re-sorts before comparing, which also passes if the outputs are in a different order.

Other example outputs tests stacking healpix and arbitrary input directories using

desi_zcatalog --add-units --recoadd-fibermap --nproc 16 \
    --group healpix \
    --indir $CFS/desi/spectro/redux/iron/healpix/sv2/dark \
    --outfile zpix-sv2-dark-pr.fits

desi_zcatalog --add-units --nproc 16 \
    --indir $CFS/desi/users/desi/test/perlmutter/ptest-250 \
    --outfile ztile-ptest-250.fits

@weaverba137 I believe this is the last PR I need before making a tag and regenerating iron redshift catalogs for more detailed data review. Please check.

[weaverba137]

I will take a look and hopefully attempt a database test-load later this week.

[weaverba137]

@sbailey, go ahead and merge when ready. I took a look at the sample files and they seem fine.