Summary

• add a server-independent JMH benchmark that invokes the existing private region-update queue through a pre-bound method handle
• model the exact 327-update representative render distribution across prompt, hand, turn-state, board, and background priorities
• verify priority/sequence execution order, processed counts, and the first-failure deferred boundary before every benchmark fork
• add a base-owned region-queue paired A/B profile with time and normalized-allocation metrics plus its workflow label

Scope

Measurement infrastructure only. No production implementation is changed, and this PR makes no optimization claim.

Validation

All compilation, contract setup, protected-benchmark smoke, and standard server.jar smoke run only on GitHub Actions. A follow-up production-only queue candidate may merge only on PASS_OPTIMIZED.