perf(memory): implement thread-local buffer pool for request bodies (#160)

[ZhuchkaTriplesix]

Closes #160

• Added src/buffer_pool.rs providing PooledBuffer with automatic RAII recycling to a thread-local buffer pool on Drop (for buffers with capacity between 16 KB and 128 KB).
• Updated run_rsgi in src/dispatch.rs to read incoming request bodies into PooledBuffer using zero-copy PyBytes::as_bytes / PyString::to_str slice extensions instead of allocating a fresh Vec<u8> on every request.
• Eliminated heap allocation and deallocation churn for request payloads on high-throughput workloads.