Closed metadata observations for an explicitly attested INSERT subset.

Exact, bounded-size snapshots of stored SQL catalog inputs.

Exact capture is accelerated by checked runtime dependency tokens. Cached
bases retain detached values only, bounded to eight entries and 32MiB of
structural weight (not exact JVM heap). Strict schemas use attribute-index
ranges; read schemas require an AEVT scan to discover undeclared catalog
attributes. The latter also visits user data.

Persistent PostgreSQL catalog identities for user-defined objects.

PostgreSQL identifies an object by (catalog relation OID, object OID),
with an optional sub-object number for columns.  Keeping that shape here
lets routines, triggers, dependencies, and future catalog work share one
transactional identity model instead of inventing per-feature IDs.