fix(memory): address review findings from Issue #88

Fixes based on multi-agent review:

Model Improvements:
- Remove duplicate index ix_procedures_agent_type (already indexed via Column)
- Fix postgresql_where to use text() instead of string literal in Fact model
- Add thread-safety to Procedure.success_rate property (snapshot values)

Data Integrity Constraints:
- Add CheckConstraint for Episode: importance_score 0-1, duration >= 0, tokens >= 0
- Add CheckConstraint for Fact: confidence 0-1
- Add CheckConstraint for Procedure: success_count >= 0, failure_count >= 0

Migration Updates:
- Add check constraints creation in upgrade()
- Add check constraints removal in downgrade()

Note: SQLAlchemy Column default=list is correct (callable factory pattern)

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

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

backend/app/models/memory/episode.py

@@ -8,6 +8,7 @@ with context, actions, outcomes, and lessons learned.
 
 from sqlalchemy import (
     BigInteger,
+    CheckConstraint,
     Column,
     DateTime,
     Enum,
@@ -119,6 +120,19 @@ class Episode(Base, UUIDMixin, TimestampMixin):
         Index("ix_episodes_project_time", "project_id", "occurred_at"),
         # For importance-based pruning
         Index("ix_episodes_importance_time", "importance_score", "occurred_at"),
+        # Data integrity constraints
+        CheckConstraint(
+            "importance_score >= 0.0 AND importance_score <= 1.0",
+            name="ck_episodes_importance_range",
+        ),
+        CheckConstraint(
+            "duration_seconds >= 0.0",
+            name="ck_episodes_duration_positive",
+        ),
+        CheckConstraint(
+            "tokens_used >= 0",
+            name="ck_episodes_tokens_positive",
+        ),
     )
 
     def __repr__(self) -> str: