sql/catalog: cache negative name lookups within a transaction

[virajchogle]

Problem

By-name resolution of an unqualified object (notably a UDT cast like 'east'::region) probes each schema on the search_path in order. The not-found result was not cached: descriptor leases carry their own embedded name, so positive lookups are transitively cached via the leased descriptor, but a negative lookup has no leasable carrier. The miss bit lives in
cachedCatalogReader.byNameState but was never exposed to the layer that decides whether to call into the lease subsystem.

Reproducer (from #171543):

CREATE TYPE region AS ENUM ('east', 'west');
SELECT x::region FROM (VALUES ('east'), ('west'), ('east'), ('west')) AS v(x);

system.namespace reads scale linearly in rows × non-matching schemas.

Fix

Add CatalogReader.IsNameKnownToNotExist, consulted in the lookupStoreCacheID layer of Collection.getNonVirtualDescriptorID. Subsequent references to the same absent name within a txn halt before
lookupLeasedID, which would otherwise call resolveName and re-read system.namespace.

Per-txn only: the cached reader is reset at txn boundaries, and in-txn creates go through lookupUncommittedID (which runs first), so a name created mid-txn is never blocked. Cross-txn negative caching is left as a follow-up.

Tests

Extends TestDataDriven in pkg/sql/catalog/internal/catkv with an is_name_known_to_not_exist directive covering post-miss / unrelated absent / positive entry / post-reset cases, mirrored across both tenant testdata files.

Fixes #171543

Epic: none

Release note: none

[blathers-crl]

Thank you for contributing to CockroachDB. Please ensure you have followed the guidelines for creating a PR.

My owl senses detect your PR is good for review. Please keep an eye out for any test failures in CI.

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is dev-inf.}