RCAL-856: WCS should roundtrip with sufficient accuracy

[nden]

Resolves RCAL-856

Closes #1272

This PR is the last one in a chain of PRs aiming to fix roundtripping of WCS transforms.
The only change here is to test data, ensuring the WCS is consistent and the inverse is wihtin the bounding box.

The order the PRs should be merged is

• Respect the bounding_box in inverse transforms gwcs#498
• AL-852: GWCS should respect the bounding box on inverse transforms stcal#263
• this one

Checklist

• added entry in CHANGES.rst under the corresponding subsection
• updated relevant tests
• updated relevant documentation
• updated relevant milestone(s)
• added relevant label(s)
• ran regression tests, post a link to the Jenkins job below. How to run regression tests on a PR

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 77.89%. Comparing base (e250097) to head (58c1477).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1273      +/-   ##
==========================================
- Coverage   78.00%   77.89%   -0.11%     
==========================================
  Files         115      115              
  Lines        7555     7563       +8     
==========================================
- Hits         5893     5891       -2     
- Misses       1662     1672      +10     

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

[ddavis-stsci]

LGTM

[schlafly]

I think I missed the part of this test that checks that things round trip?

This WCS contains only the v2v3tosky bit, and not the CRDS distortion bit. That's good since it makes for an easier unit test but bad since it won't catch the only case we have had problems here so far, where the failure to round trip was in the CRDS distortion file. Do we intend to check the CRDS reference files as well?

I note that the bounding box is inconsistent with the size of the image, and that we're pretending that the sizes of the pixels are ~1". I think that's likely fine. romanisim has some code for making up a "fake" distortion function here
https://github.com/spacetelescope/romanisim/blob/main/romanisim/tests/test_wcs.py#L19-L38
and using that in unit tests instead of the actual CRDS distortion functions. It's not obvious to me that that is helpful here but I link it in case you would have rather used some kind of distortion function.

[schlafly]

@nden , do you plan to go forward with this in the future, or should we merge now?

[nden]

This requires a change in gwcs which I haven't had time to complete. I'll take it out of Draft when it's ready.