feat(memory): implement memory reflection service (#99)

Add reflection layer for memory system with pattern detection, success/failure
factor analysis, anomaly detection, and insights generation. Enables agents to
learn from past experiences and identify optimization opportunities.

Key components:
- Pattern detection: recurring success/failure, action sequences, temporal, efficiency
- Factor analysis: action, context, timing, resource, preceding state factors
- Anomaly detection: unusual duration, token usage, failure rates, action patterns
- Insight generation: optimization, warning, learning, recommendation, trend insights

Also fixes pre-existing timezone issues in test_types.py (datetime.now() -> datetime.now(UTC)).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

backend/app/services/memory/__init__.py

@@ -90,6 +90,9 @@ from .types import (
     WorkingMemoryItem,
 )
 
+# Reflection (lazy import available)
+# Import directly: from app.services.memory.reflection import MemoryReflection
+
 __all__ = [
     "CheckpointError",
     "ConsolidationStatus",