Streamline prod zcatalog creation

[sbailey]

For Kibo, the redshift catalogs were generated by hand by cutting and pasting commands like

salloc -N 1 -C cpu -A desi -t 04:00:00 -q interactive
cd $DESI_SPECTRO_REDUX/$SPECPROD

SURVEY=cmx
PROGRAM=other
desi_zcatalog -i /dvs_ro/cfs/cdirs/desi/spectro/redux/$SPECPROD/healpix/$SURVEY/$PROGRAM/ \
              -o zcatalog/v1/zpix-$SURVEY-$PROGRAM.fits -g healpix --nproc 64 |& tee zcatalog/v1/logs/zpix-$SURVEY-$PROGRAM.log

SURVEY=sv1
PROGRAM=other
desi_zcatalog -i /dvs_ro/cfs/cdirs/desi/spectro/redux/$SPECPROD/healpix/$SURVEY/$PROGRAM/ \
              -o zcatalog/v1/zpix-$SURVEY-$PROGRAM.fits -g healpix --nproc 64 |& tee zcatalog/v1/logs/zpix-$SURVEY-$PROGRAM.log

for SURVEY in sv1 sv2 sv3 main special; do
  for PROGRAM in dark bright backup; do
    echo ----- $SURVEY $PROGRAM -----
    nohup desi_zcatalog -i /dvs_ro/cfs/cdirs/desi/spectro/redux/$SPECPROD/healpix/$SURVEY/$PROGRAM/ \
                        -o zcatalog/v1/zpix-$SURVEY-$PROGRAM.fits -g healpix --nproc 64 &>> zcatalog/v1/logs/zpix-$SURVEY-$PROGRAM.log
  done
done

The particular combination of SURVEY,PROGRAM was derived from what exists in the tiles-kibo.csv file and then typed into bash by hand. This could be standardized and streamlined into a single job that reads tiles-kibo.csv to know what to do and generate any missing zcatalog files.

[sbailey]

Additionally, Kibo required post-facto adding of units and column descriptions with annotate_fits -C $DESIDATAMODEL/py/desidatamodel/data/column_descriptions.csv ... . Make sure that is included in the wrapper, or otherwise automatically part of desi_zcatalog.

[sbailey]

Now that @stephjuneau in desihub/desidatamodel#196 provided column descriptions that are short enough to be useful for comments, desispec.scripts.zcatalog could start propagating those comments into write_bintable. Thanks to @weaverba137 for noticing that zcatalog could be passing comments but wasn't.

[weaverba137]

I've already incorporated the change into the read-zbest branch. Note that there are still a few cases where the column name is so long that the short description is still truncated, such as PSF_TO_FIBER_SPECFLUX or, especially, MEAN_PSF_TO_FIBER_SPECFLUX.

I've also found some columns that are missing descriptions:

• FIRSTNIGHT
• NUMEXP
• NUMTILE
• MIN_MJD
• MAX_MJD
• MEAN_MJD
• IN_COADD_B
• IN_COADD_R
• IN_COADD_Z