restore column units on cast

[braingram]

Resolves JP-3482

Closes [spacetelescope/jwst#8109]

When a DataModel containing a table extension is loaded from a FITS file, stdatamodels currently discards the units defined in the TUNITS headers. This is due to a _cast from a FITS_rec to a ndarray (to protect against the numerous gotchas in FITS_rec):

stdatamodels/src/stdatamodels/properties.py

Line 86 in b7eada9

val = util.gentle_asarray(val, dtype, allow_extra_columns=allow_extra_columns)

then a conversion back to a FITS_rec to allow jwst code to continue to interact with a FITS_rec:

stdatamodels/src/stdatamodels/properties.py

Line 89 in b7eada9

val = _as_fitsrec(val)

This PR is an alternative to #238 and a stop-gap measure until #240 can be resolved.

Checklist

• added entry in CHANGES.rst (either in Bug Fixes or Changes to API)
• updated relevant tests
• updated relevant documentation
• updated relevant milestone(s)
• added relevant label(s)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (bbc9016) 64.78% compared to head (3dd1cfd) 64.82%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #242      +/-   ##
==========================================
+ Coverage   64.78%   64.82%   +0.04%     
==========================================
  Files         103      103              
  Lines        5688     5695       +7     
==========================================
+ Hits         3685     3692       +7     
  Misses       2003     2003              

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

[braingram]

Regression tests started at: https://plwishmaster.stsci.edu:8081/job/RT/job/JWST-Developers-Pull-Requests/1084/
the above tests crashed and burned due to many artifactory connection errors
I restarted them with the hope that this was transient:
https://plwishmaster.stsci.edu:8081/job/RT/job/JWST-Developers-Pull-Requests/1085/
The re-run shows 30 errors. None of which on the surface appear related to these changes. All contain the following:

E              Extra column R_MIREMI of format D in a
E              Extra column S_MIREMI of format D in a

which points towards truth files that haven't been updated yet for #200

[braingram]

@jemorrison I can't seem to request you as a reviewer but thanks again for testing out this PR and confirming that it fixed your issue. Please feel free to review/comment on the PR if you have a chance to give it a look.

[jemorrison]

Looks reasonable to me and it works.

[hbushouse]

LGTM

[hbushouse]

There have been a couple rounds of regtest truth file updates over the past few days, so it might be worthwhile to run this PR branch again, one last time.

[hbushouse]

OK, now that this has been merged, I'll just kick off a regular RT run, which will use stdatamodels/master.

[braingram]

Thanks! Sorry for the quick draw on the merge.

[hbushouse]

Regtest run https://plwishmaster.stsci.edu:8081/job/RT/job/JWST/2686/ shows only unrelated failures/errors. So this looks good (at least it does no obvious harm).