feat(policy): add opt-in composite reward scoring with audit tests

policy/context.py

@@ -12,6 +12,7 @@ class Decision:
     reason: str = ""
     source: str = "code"          # code | jev | fallback
     confidence: float | None = None
+    scoring: dict | None = None
 
     def __str__(self) -> str:
         params = ", ".join(f"{k}={v!r}" for k, v in self.params.items())