feat: preserve governed OpenAI tool messages in 1.4.0

Add faithful single-choice chat results, tool stream activity, error-frame retries, and isolated protocol caching.

Reject empty cached completions with the same value predicate as transport, add regression coverage, and adopt Apache-2.0.

tests/unit/test_retry.py

@@ -90,6 +90,7 @@ class FakeTransport:
         reasoning_effort,
         first_token_event,
         content_event,
+        response_mode,
     ):
         self.calls.append((source.name, call_id))
         self.efforts.append(reasoning_effort)
@@ -132,6 +133,7 @@ class _GenClockTransport:
         reasoning_effort,
         first_token_event,
         content_event,
+        response_mode,
     ):
         advance = self._advances.pop(0)
         stall = self._stalls.pop(0) if self._stalls is not None else None
@@ -144,6 +146,7 @@ class _GenClockTransport:
             reasoning_effort=reasoning_effort,
             first_token_event=first_token_event,
             content_event=content_event,
+            response_mode=response_mode,
         )
         if stall is not None:
             result = dataclasses.replace(result, stall_ms=stall[0], stall_count=stall[1])