test(bench): add task ownership A/B profile

[LimiNode]

Summary

• compare std::function construction/copy with payload ownership
• measure TaskExecutor batch admission versus one-at-a-time round-trip
• measure direct MpscRingAny publication for std::function and uint64 payloads
• add bounded round-trip sample count and Linux C++17 smoke coverage

Verification

• MSVC Release C++17 with MPSC ring: 100k batch/ring samples, bounded 1k round-trip samples
• MSVC Release C++17 with mutex deque: smoke run
• all changes are benchmark/CI/docs only; production implementation unchanged