search-with-proof: name the exhausted limit, keep partial evidence, tighter reservation

budget_spent carries end_detail (spend, task_spend, reason_rounds, ...); a task
whose allowance cannot cover the next call finishes with its facts; per-task
metrics in pilot records; input reservation bytes/2 + 1024 per message (measured
3-4x over-reservation before).

Co-Authored-By: Claude Fable 5.1 <[email protected]>

src/openultrasast/plane/budget.py

@@ -23,8 +23,9 @@ _USAGE_FIELDS = ("prompt_tokens", "prompt_cache_hit_tokens", "completion_tokens"
 class BudgetExhausted(RuntimeError):
     """Raised before a call would start once the usd or calls ceiling is reached."""
 
-    def __init__(self, message: str, *, usd: float | None, calls: int) -> None:
-        super().__init__(message)
+    def __init__(self, message: str, *, usd: float | None, calls: int, end_detail: str = "spend") -> None:
+        super().__init__(f"{end_detail}: {message}")
+        self.end_detail = end_detail
         self.usd = usd
         self.calls = calls