update validredshifts.validate() to work on SV observations

[moustakas]

The convenience script validredshifts.validate() implements the nominal criteria to select "good" redshifts; however, it currently only works on main-survey targets. This simple PR enhances the function to work on SV (but not CMX) data.

This script is not used in any part of production so I have not taken the time to write unit tests, but it would be helpful if this could be merged eventually as part of the template work I'm doing.

[sbailey]

The code looks fine, but I haven't been able to reproduce the failing case on main so I'm not sure I'm evaluating the right thing. e.g. using desispec/main (not this PR branch) to validate sv1 tile 80611:

from desispec.validredshifts import validate
validate('/global/cfs/cdirs/desi/spectro/redux/iron/tiles/cumulative/80611/20201222/redrock-0-80611-thru20201222.fits')

-->
<Table length=500>
GOOD_BGS GOOD_LRG GOOD_ELG GOOD_QSO      TARGETID                 Z            ZWARN COADD_FIBERSTATUS
  bool     bool     bool     bool         int64                float64         int64       int32      
-------- -------- -------- -------- ------------------ ----------------------- ----- -----------------
    True     True     True    False  39628462817808192      0.1495700223052229     0                 0
   False    False    False    False 616089215125422495      1.5405229350632148     1                 0
    True     True     True    False  39628462822001504     0.10297132568175493     0                 0
    True     True     True    False  39628468098437649      0.3301800074160998     0                 0
...

Regarding "it currently only works on main-survey targets" -- does it crash, or give the wrong answer, or only fail on some tiles, or?

[moustakas]

@sbailey thanks for the review. See these lines:
https://github.com/desihub/desispec/pull/2122/files#diff-00224424e1a4095bc1c7f95139478923c6ac2961fc4e33106de8455e262621d0L117-L119

If you set return_target_columns=True you'll see that it only returns the main-survey target columns (DESI_TARGET, BGS_TARGET, etc.).

In this PR I use desitarget.targets.main_cmx_or_sv (@geordie666) to determine which targeting bits to return, which in the case of your example would be SV1_DESI_TARGET, SV1_BGS_TARGET, etc.

[sbailey]

Clarifying my confusion: The issue isn't a crash, or that desispec.validredshifts.validate itself was returning DESI_TARGET vs. SV1_DESI_TARGET, etc., but rather that internally validate was using the wrong target columns when calculating the return columns like ELG, LRG, QSO, BGS_ANY, etc. and thus getting them wrong (usually False).

Thanks for the update. Merging now.