Add comprehensive benchmarks for hot paths with allocation tracking

[jgowdy-godaddy]

Summary

This PR implements comprehensive performance benchmarks for Asherah's critical hot paths with detailed memory allocation tracking. The benchmarks are designed to detect performance regressions and guide optimization efforts for production deployments.

Hot Path Benchmarks Added

• SessionFactory.GetSession: Both cached and uncached scenarios
• Session.Encrypt/Decrypt: Core operations with 1KB realistic payloads
• Round-trip operations: Full encrypt-decrypt cycles
• Key cache operations: GetOrLoad and GetOrLoadLatest performance
• Reference counting: Cached crypto key lifecycle operations
• Memory pressure testing: Large payload operations (64KB)

Concurrent Benchmarks Added

• Concurrent encryption/decryption: Performance under load
• Concurrent session creation: SessionFactory scalability
• Concurrent key cache access: Both same-key and unique-key scenarios
• Mixed operations: Real-world encrypt/decrypt patterns
• Session cache behavior: Cache performance under concurrent access
• Reference counting concurrency: Thread-safe reference management

Key Features

✅ Allocation tracking: All benchmarks use b.ReportAllocs() for detailed memory profiling
✅ Realistic workloads: 1KB standard payloads, 64KB for memory pressure testing
✅ Import cycle safe: Custom minimal mock implementations avoid dependency issues
✅ Production patterns: Benchmarks mirror real-world usage scenarios
✅ Concurrent testing: Both single-threaded and multi-threaded workloads

Sample Results

BenchmarkSessionFactory_GetSession_HotPath-16      6408549    180.3 ns/op    472 B/op   10 allocs/op
BenchmarkSessionFactory_GetSession_Cached-16      53285770     22.46 ns/op      0 B/op    0 allocs/op
BenchmarkSession_Encrypt_HotPath-16                  25484     46967 ns/op   2210 B/op   27 allocs/op
BenchmarkSession_Decrypt_HotPath-16                1232919       988.1 ns/op    256 B/op    8 allocs/op
BenchmarkSession_EncryptDecrypt_RoundTrip-16         24415     49267 ns/op   3490 B/op   36 allocs/op
BenchmarkKeyCache_GetOrLoad_WithAllocation-16     17500576        67.21 ns/op     40 B/op    2 allocs/op

Testing

• ✅ All existing tests pass
• ✅ New benchmarks execute successfully
• ✅ Memory allocation tracking confirmed working
• ✅ Both sequential and concurrent benchmarks validated

Test plan

• Run go test -bench=Benchmark -benchtime=1s to verify all benchmarks execute
• Run go test ./... to ensure no regressions
• Verify allocation tracking reports meaningful data
• Test both hot path and concurrent benchmark suites

Generated with Claude Code

[aka-bo]

Great work on the Session benchmarks! Those are valuable additions.

However, there's significant overlap with existing benchmarks. The project already has omprehensive key cache benchmarks in key_cache_benchmark_test.go that cover the same GetOrLoad/GetOrLoadLatest scenarios you're testing. The duplicated benchmarks should be removed.

Also, ReportAllocs() just enables allocation tracking by default (same as running existing benchmarks with -benchmem). Consider adding -benchmem to ./build/go/benchmark_test.sh instead - that would give all benchmarks allocation tracking without code changes.

The file naming should follow the project's established {thing-to-benchmark}_benchmark_test.go convention. Consider session_benchmark_test.go to match the existing pattern.

The .tool-versions file needs to be removed from the PR too.