Remove the ScarletModelData to LsstScarletModelData converter (DM-55150)

[fred3m]

A base ScarletModelData cannot be promoted to an LsstScarletModelData: the latter requires a single catalog-wide observed PSF and band list, while a ScarletModelData carries them per blend. The conversion never produced a usable model in the science pipelines, so drop the storage-class converters entry (and the now-removed scarlet_model_to_lsst_scarlet_model it referenced, deleted in a companion commit on this ticket).

Old ScarletModelData datasets continue to load at their own storage class.

Checklist

• ran Jenkins
• added a release note for user-visible changes to doc/changes
• (if changing dimensions.yaml) make a copy of the updated dimensions.yaml in configs/old_dimensions and update the list in doc/lsst.daf.butler/dimensions.rst

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.89%. Comparing base (78e7f29) to head (eb59e68).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1386      +/-   ##
==========================================
- Coverage   89.90%   89.89%   -0.01%     
==========================================
  Files         373      373              
  Lines       51089    51089              
  Branches     5953     5953              
==========================================
- Hits        45930    45929       -1     
  Misses       3766     3766              
- Partials     1393     1394       +1     

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

[timj]

@TallJimbo I am guessing we need to keep this converter around for a while given #1411

[TallJimbo]

I'm not sure; #1411 is about the other direction, which in practice is about supporting new pipelines/tasks on old repos. That's important because most extant repos qualify as "old" - i.e. they have the old storage class for the dataset type where this is used.

So I think this PR is just for supporting old pipelines/tasks (which now just means older software versions) against new repos (dp2* at USDF). And that means it can probably follow a regular-software deprecation cycle.

[TallJimbo]

Oh, no, nevermind: we do need this in the long term, because if #1411 makes a new PipelineTask write the old storage class in an old repo, a consuming new PipelineTask will then need the conversion provided by this PR. Which, I suppose, might not work given the limitations that motivated this PR. So it's possible this whole scheme is doomed to failure, and we will have to do something else like a dataset type rename for backwards compatibility.