feat(memory): add semantic memory implementation (Issue #91)

Implements semantic memory with fact storage, retrieval, and verification:

Core functionality:
- SemanticMemory class for fact storage/retrieval
- Fact storage as subject-predicate-object triples
- Duplicate detection with reinforcement
- Semantic search with text-based fallback
- Entity-based retrieval
- Confidence scoring and decay
- Conflict resolution

Supporting modules:
- FactExtractor: Pattern-based fact extraction from episodes
- FactVerifier: Contradiction detection and reliability scoring

Test coverage:
- 47 unit tests covering all modules
- extraction.py: 99% coverage
- verification.py: 95% coverage
- memory.py: 78% coverage

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

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

backend/app/services/memory/semantic/__init__.py

@@ -1,3 +1,4 @@
+# app/services/memory/semantic/__init__.py
 """
 Semantic Memory
 
@@ -5,4 +6,22 @@ Fact storage with triple format (subject, predicate, object)
 and semantic search capabilities.
 """
 
-# Will be populated in #91
+from .extraction import (
+    ExtractedFact,
+    ExtractionContext,
+    FactExtractor,
+    get_fact_extractor,
+)
+from .memory import SemanticMemory
+from .verification import FactConflict, FactVerifier, VerificationResult
+
+__all__ = [
+    "ExtractedFact",
+    "ExtractionContext",
+    "FactConflict",
+    "FactExtractor",
+    "FactVerifier",
+    "SemanticMemory",
+    "VerificationResult",
+    "get_fact_extractor",
+]