Update absolute separation default.

[schlafly]

This PR updates the default absolute separation used in tweakreg in romancal, as discussed in #1329.

Closes #1329.

If multiple sources within 'tolerance' can be found, tweakreg may fail. Accordingly, 'separation' should always be larger than 'tolerance.' The default for tweakreg were not consistent with this requirement, which led to a failure to tweak the WCS in some cases. This PR updates the default to avoid this case.

Regression tests fail due to very small differences in the stars that tweakreg selects for matching. This strikes me as okay. There's room in the future to choose better values for these parameters; they should by a few times larger than the expected uncertainty in the pointing. But this resolves the crashes @braingram was seeing.

Checklist

• added entry in CHANGES.rst under the corresponding subsection
• updated relevant milestone(s)
• ran regression tests: https://github.com/spacetelescope/RegressionTests/actions/runs/10251136980

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 78.57%. Comparing base (3189a0a) to head (f180f84).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1352   +/-   ##
=======================================
  Coverage   78.57%   78.57%           
=======================================
  Files         117      117           
  Lines        7861     7861           
=======================================
  Hits         6177     6177           
  Misses       1684     1684           

Flag	Coverage Δ		*Carryforward flag
nightly	62.26% <ø> (ø)		Carriedforward from 3189a0a

*This pull request uses carry forward flags. Click here to find out more.

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

[schlafly]

@ddavis-stsci , mind reviewing? I think we want this to go into B15 so that tweakreg doesn't crash on some SCAs.

[braingram]

LGTM. Adding a check similar to what was added in spacetelescope/jwst#8476 might make sense as these are user (and pars file) controllable parameters. Having an informative error seems more useful than the step crashing. I don't have any objection to adding this check in a different PR but I suggest we keep the issue open if that's the plan (or make a new one).

[braingram]

Ignore my comment about adding the check. Once this step uses stcal it will inherit the check:
https://github.com/spacetelescope/stcal/blob/809de2398dc3843821d71fc5cba88ae591a16eb9/src/stcal/tweakreg/tweakreg.py#L129

[schlafly]

Perfect. Yeah, I agree that this kind of check makes more sense downstream (really in tweakreg itself?) than in romancal / jwst.

[schlafly]

I okified the regtests and am merging.