benchmarking

std::shared_mutex lost to std::mutex by 5.5x. High-performance LOB engine challenging lock assumptions.

Worker pool tuning, grid partitioning, pub/sub architecture, and CPU profiling. Side-by-side benchmark of shared-memory vs networked concurrency.