fix(agent): deliver LLM errors to streaming channels and avoid polluting session context

When the LLM returns an error (e.g. 429 quota exceeded, stream timeout),
streaming channels silently drop the error message because `_streamed=True`
is set in metadata even though no content was actually streamed.

This change:
- Skips setting `_streamed` when stop_reason is "error", so error messages
  go through the normal channel.send() path and reach the user
- Stops appending error content to session history, preventing error
  messages from polluting subsequent conversation context
- Exposes stop_reason from _run_agent_loop to enable the above check

tests/tools/test_message_tool_suppress.py

@@ -107,7 +107,7 @@ class TestMessageToolSuppressLogic:
         async def on_progress(content: str, *, tool_hint: bool = False) -> None:
             progress.append((content, tool_hint))
 
-        final_content, _, _ = await loop._run_agent_loop([], on_progress=on_progress)
+        final_content, _, _, _ = await loop._run_agent_loop([], on_progress=on_progress)
 
         assert final_content == "Done"
         assert progress == [