Tests: inc-site attribute extension construct path + memoized reference target

[rly]

Summary

Closes the test gaps PR #283 flagged but never landed:

• Construct (read) round-trip for inc-site added attributes — TestBuildGroupAddedAttr.test_construct_added_attr and TestBuildDatasetAddedAttrs.test_construct_added_attr build a holder containing an extended Bar / BarData with an inc-site ext_attr, then construct from the resulting builder and assert the attribute round-trips. Required adding @ObjectMapper.constructor_arg('ext_attr') to ExtBarMapper / ExtBarDataMapper so the inner mapper can recover an inc-site attribute that is not in its def-site spec.
• Reference + memoization positive test — TestBuildDatasetAddedAttrsWithRef.test_build_added_attr_with_ref builds a holder where the direct extended bar_data field and a sibling typed reference dataset (BarRefDataset) both point at the same BarData instance, and asserts (a) the BarData builder carries the inc-site ext_attr and (b) the reference dataset's ReferenceBuilder resolves to that same memoized builder. Modern code only had the negative-case TestBuildDatasetOfReferencesUnbuiltTarget.

Also uncomments TestBuildGroupRefinedAttr.test_build_refined_attr_wrong_type (the attribute counterpart of TestBuildDatasetRefinedDtype.test_build_refined_dtype_convert — verifies build silently converts int -> int64 when the inc-site spec refines the dtype). The inline TODO is preserved noting that ideally a DtypeConversionWarning would be emitted, mirroring the dataset case.

The four touched mixin / test setUps were refactored to use tests.unit.helpers.utils.create_test_type_map instead of hand-rolling SpecCatalog / SpecNamespace / NamespaceCatalog / TypeMap.

Notes on design choices

PR #283 used a named-untyped dataset with dtype=RefSpec directly holding a Container value. Modern convert_dtype requires a ReferenceBuilder for that path, so the with-ref test instead uses a typed Data subclass (BarRefDataset) holding object references — the canonical "container of references" pattern (mirrors Baz in test_io_map_data.py).

These tests do not move statement / branch coverage numbers (the underlying code paths are already exercised by other tests). They add behavioral guarantees at the inc-site-extension boundary that PR #283 was worried about.

Test plan

• pytest tests/unit/build_tests/mapper_tests/test_build.py -> 20 passed
• pytest tests/ -> 1891 passed, 22 skipped, 1 xfailed

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.20%. Comparing base (41b7f08) to head (234d1c9).

Additional details and impacted files

@@           Coverage Diff           @@
##              dev    #1468   +/-   ##
=======================================
  Coverage   93.20%   93.20%           
=======================================
  Files          41       41           
  Lines       10176    10176           
  Branches     2103     2103           
=======================================
  Hits         9485     9485           
  Misses        415      415           
  Partials      276      276           

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.