feat(observer): add performance test for snapshot serialization with large metric sets

[ProtocolWarden]

Summary

Add comprehensive performance tests for snapshot serialization with large metric sets to verify serialization efficiency across different data volumes and signal combinations.

This PR implements 24 new performance tests that measure:

• Serialization latency across JSON, JSONL, and YAML formats
• Memory efficiency under large metric loads (100 → 5K → 50K tests)
• Throughput metrics and scalability characteristics
• Format comparison (size vs. speed tradeoffs)
• Roundtrip consistency (serialize → deserialize verification)

Changes

New Test Suite

• 24 new performance tests in TestSnapshotSerializationLargeMetrics class
• 3 performance tiers: SMALL (100 tests), MEDIUM (5K tests), LARGE (50K tests)
• Enhanced snapshot factory: create_large_snapshot(tier, index, seed) for reproducible test data
• 7 helper functions for realistic data generation across all signal types

Performance Assertions Verified

All performance thresholds met across formats and tiers:

• JSON: <50ms (small), <500ms (medium), <5s (large) ✅
• JSONL: <10ms (small), <50ms (medium), <500ms (large) ✅
• YAML: <100ms (small), <1s (medium), <10s (large) ✅
• Memory: <10MB (small), <100MB (medium), <500MB (large) ✅
• Scalability: Linear growth validated across all formats ✅

Test Results

• ✅ 24 new tests: all PASSING
• ✅ 37 total performance tests: all PASSING
• ✅ 1,141 observer unit tests: all PASSING
• ✅ 7,373 total tests: all PASSING (7,195 unit + 178 integration)
• ✅ Code quality: Ruff 0 violations, Custodian 0 findings
• ✅ No regressions detected

🤖 Generated with Claude Code

[ProtocolWarden]

Self-review concerns — auto-fixing (up to 6 attempts; re-queued if still unresolved):

Diff truncation prevents full verification. The visible ~50% of the test class implementation shows high quality: helper functions are well-designed, create_large_snapshot factory covers 3 tiers comprehensively, 13+ visible tests follow consistent patterns with appropriate performance thresholds, imports are correct (CheckSignal→TestSignal migration consistent), and no code quality issues detected. However, the diff cuts off mid-implementation at 60000 characters, obscuring ~11 remaining tests and the complete test class definition. PR documentation claims 24 tests implemented with all tests passing locally, but this cannot be verified from the truncated diff. To complete review: (1) provide full diff showing all 24 test methods, or (2) run tests locally to confirm reported passing results align with full implementation.