fix(ai): admit bounded warnings while resuming

[first-assist]

Outcome

The strict Codex app-server actor now admits the documented generic warning notification during an exact retained thread/resume handshake, in addition to the existing correlated-turn window.

Warning text is validated and discarded. Admission remains thread-correlated, count- and byte-bounded, and cannot advance resume readiness. New-thread creation, deletion, state-index scans, idle periods, malformed frames, mismatched thread IDs, and flooding remain rejected.

Compatibility

• graphql-orm-ai: 0.95.14
• AI schema module: unchanged at 0.64.0
• no database, GraphQL SDL, protected-payload, backup, restore, or data migration
• hosts should continue reading after RuntimeWarning until exact resume readiness is proven

Verification

• focused warning and resume interleaving tests
• full provider-enabled graphql-orm-ai unit, integration, and doc tests
• warnings-denied Clippy and Rustdoc
• PascalCase contract and documentation
• PostgreSQL and MSSQL compile-only lanes
• documentation, workspace dependency, release state, package release, deterministic manifest, formatting, diff, and SemVer gates

Documentation impact

• Documentation updated

README, changelog, migration guide, implementation status, and generated workspace inventory are updated.