test(parity): compare against oold-js instead of oold-schema's scripts

[simontaurus]

Closes #136. Pairs with OO-LD/oold-schema#156, which removes the scripts this stops pointing at.

validate.mjs, schema_to_frame.mjs and pattern_lint.mjs now live in OO-LD/oold-js, extracted unchanged and tagged v0.1.0.

Change

• New reference_dir fixture reading OOLD_JS_DIR, alongside upstream_dir for OOLD_SCHEMA_DIR. The corpus and the reference are different repositories now and a machine can have either without the other.
• A reference fixture runs the validator and both subprocess call sites collapse into it. It passes --meta <corpus>/meta, so both implementations read one file set and a disagreement can only come from the code - oold-js vendors no meta-schemas.
• CI clones oold-js at a pinned tag and npm cis it. The corpus stays on main: pinning the reference makes a change to it deliberate, while leaving the corpus moving is what lets parity notice a specification change the fixture snapshot cannot.
• CLAUDE.md, docs/architecture.md and docs/how-to/validation.md updated; the pin they cited (validate.mjs at v1.0.0-rc.2) no longer exists.

Without OOLD_JS_DIR the two verdict-comparing tests skip and the four corpus-only tests still run.

Verification

• both variables set: 6 passed
• OOLD_SCHEMA_DIR only: 4 passed, 2 skipped, naming OOLD_JS_DIR in the skip reason
• pytest: 565 passed, 9 skipped
• make docs-test: no issues

[github-actions]

Release preview

No version bump from the current commits (stays at v0.18.2). Use conventional commit types (feat, fix, ...) to trigger a release.

Changelog preview (truncated)

_{Preview via python-semantic-release and conventional commits.}

[github-actions]

📊 Benchmark Results

Click to see benchmark comparison

📊 Benchmark Comparison (threshold: 1.3x)
============================================================

➖ Unchanged (within threshold):
  ➖ test_simple_dict_document_store: 0.0017s → 0.0017s (+1.5%)
  ➖ test_sqlite_document_store: 0.0019s → 0.0019s (-0.6%)
  ➖ test_local_sparql_store: 0.0385s → 0.0399s (+3.6%)
  ➖ test_oneof_subschema: 0.0594s → 0.0597s (+0.4%)
  ➖ test_enum_docstrings: 0.0508s → 0.0498s (-1.9%)
  ➖ test_subclass_inheritance: 0.0560s → 0.0550s (-1.7%)
  ➖ test_class_hierarchy: 0.0536s → 0.0508s (-5.2%)
  ➖ test_core[v1]: 0.0388s → 0.0400s (+3.1%)
  ➖ test_core[v2]: 0.0445s → 0.0443s (-0.6%)
  ➖ test_schema_generation[v1]: 0.0017s → 0.0017s (-0.4%)
  ➖ test_schema_generation[v2]: 0.0028s → 0.0028s (+0.4%)
  ➖ test_simple_json: 0.0007s → 0.0007s (+0.4%)
  ➖ test_complex_graph: 0.0016s → 0.0016s (-1.9%)

============================================================
Summary: 0 regressions, 0 improvements, 13 unchanged
============================================================

✅ No significant performance regressions

Threshold: 1.3x (30% slower triggers a regression warning)

Note: Benchmarks are informational only and won't fail the build.

_{💡 Tip: Download the benchmark-results artifact for detailed JSON data}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!