test: refresh PostgreSQL coverage after native grammar expansion

Regenerate the pinned PostgreSQL 18.4 correctness corpus and selected CI cases
against implementation commit 5216fce. The 51,415-statement inventory now has
43,373 deep parses (84.4%), up from 27,014, for a gain of 16,359 statements.
All 1,274 previously classification-only statements now have complete ASTs.

Compare every refreshed result with the independently replayed inventory and
confirm zero baseline regressions. Review 5,919 transitions between existing
unsupported categories: 5,165 trailing-input to error, 723 type-mismatch to
error and 31 trailing-input to partial. These are stricter unsupported grammar
outcomes, not losses of accepted statements.

Refresh the generated compatibility report and 10,983 selected CI fixtures.
Keep PostgreSQL 17.7/18.4 oracle pins unchanged. The report identifies the code
commit measured; these artifacts record correctness and grammar coverage,
not benchmark results or a claim of complete PostgreSQL grammar parity.

Validation: the full pinned compatibility workflow, statement-type mapping
tests and all 87 Python harness tests pass against the refreshed baseline.
No benchmark code, reports or benchmark-only documentation changes are included.

Author: renecannao