feat: implement signature caching to reduce CPU usage

[okdas]

Summary

This PR implements a signature caching system to address the CPU bottleneck identified in pprof analysis where ring signature operations were consuming 50% of CPU time.

Problem

• Ring signature cryptographic operations consuming 50% of CPU time
• Each signature computation takes 10-50ms
• Many requests within a session are identical (same payload, supplier, app)
• No caching mechanism existed, causing redundant expensive computations

Solution

Implemented a comprehensive signature caching system with:

• LRU cache with 100k entry capacity
• 15-minute TTL matching session duration
• Thread-safe with in-flight computation tracking
• Prevents duplicate work when multiple goroutines request same signature

Key Components

1. Signature Cache Implementation

• protocol/shannon/signature_cache.go: Core caching logic
• Cache key: SessionID + SupplierAddr + AppAddr + PayloadHash
• Handles concurrent access with sync.Map for in-flight tracking
• Automatic TTL-based expiration and cleanup

2. Integration with Signer

• Modified protocol/shannon/signer.go to use cache
• Transparent integration - falls back to computation if caching fails
• Cache statistics available for monitoring

3. Prometheus Metrics

• shannon_signature_cache_hits_total: Cache hit counter
• shannon_signature_cache_misses_total: Cache miss counter with reasons
• shannon_signature_cache_size: Current cache size gauge
• shannon_signature_cache_evictions_total: Eviction counter by reason
• shannon_signature_cache_compute_time_seconds: Computation time histogram

4. Comprehensive Testing

• Unit tests for all cache behaviors
• Concurrent access stress tests
• TTL expiration tests
• Performance benchmarks

Memory Usage

• ~50-55MB at full capacity (100k entries)
• Each entry: ~500 bytes (key + signature + metadata)

Testing

• ✅ All unit tests pass
• ✅ No regressions in existing tests
• ✅ Linting passes

Next Steps (Future PRs)

• Add YAML configuration for cache parameters
• Allow per-service cache enable/disable
• Monitor production metrics to tune cache size and TTL

Review Notes

This is a draft PR for initial review. The implementation is complete and tested. We should monitor cache hit rates in production to determine actual effectiveness.