feat: upgrade ring buffer indices to std::atomic

[HawkinWay]

Changes

• Replace non-atomic read/write indices with std::atomic
• Add atomic synchronization for producer/consumer coordination
• Update benchmark infrastructure for comparing implementations
• Add seq_cst benchmark results

Benchmark

Before (non-atomic): ~313M ops/s

After (default std::atomic seq_cst): ~60M ops/s

Notes

This implementation uses the default memory_order_seq_cst.
Future improvements will explore acquire/release ordering and cache-line optimization.