DM-45119: Update logic for checking overlap dimensions in relations

[andy-slac]

New logic only checks for required elements instead of full set of
dimensions. This fixes special case of healpix11 dataset types
in dc2 repo. There is no unit test for this special case, but I verified
that query-datasets and query-data-ids now work correctly.

Checklist

• ran Jenkins
• added a release note for user-visible changes to doc/changes
• (if changing dimensions.yaml) make a copy of dimensions.yaml in configs/old_dimensions

[codecov]

Codecov Report

Attention: Patch coverage is 90.90909% with 1 line in your changes missing coverage. Please review.

Project coverage is 89.37%. Comparing base (c108883) to head (2111e21).

Files	Patch %	Lines
...thon/lsst/daf/butler/registry/dimensions/static.py	75.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1043      +/-   ##
==========================================
- Coverage   89.38%   89.37%   -0.01%     
==========================================
  Files         359      359              
  Lines       45642    45647       +5     
  Branches     9357     9360       +3     
==========================================
+ Hits        40795    40799       +4     
  Misses       3521     3521              
- Partials     1326     1327       +1     

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

[TallJimbo]

I've made a suggestion for how to make this a bit more general, but overall I'm very much in favor of doing just this much and no more on the old query system.

[TallJimbo]

I think we can make this a little safer by keeping the old check and the new one, and combinations between them:

group1 = self.universe[element1].minimal_group
group2 = self.universe[element2].minimal_group
overlap_relationships = {
    frozenset(a | b) for a, b in itertools.product(
        [group1.names, group1.required],
        [group2.names, group2.required],
    )
}
if not overlap_relationships.isdisjoint(existing_relationships):
    return context.preferred_engine.make_join_identity_relation(), False

I'm particularly interested in keeping the old check, because while the required dimensions are the ones brought in by a dataset join, the full dimensions are the ones brought in by the temp tables created by materialize, and I'm not sure we have enough test coverage of QG generation edge cases to spot any regressions this change might introduce there otherwise.