perf: add benchmark profiling mode and fix quadratic repr bottleneck

Add --profile flag to bench/__main__.py that bypasses pyperf and runs
each benchmark under cProfile, printing per-benchmark hotspot summaries
and writing .prof files for later analysis. Add benchmark-profile and
benchmark-profile-spy Makefile targets.

Fix O(n^2) performance regression in _get_model_input_state where
f-strings eagerly evaluated repr(state) on every call, triggering
pydantic __repr__ across all accumulated messages. Move error message
construction into the error path so repr is only called when needed.
This yields a 3-5x speedup on react_agent_100x benchmarks.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

libs/prebuilt/langgraph/prebuilt/chat_agent_executor.py

@@ -638,15 +638,18 @@ def create_react_agent(
             messages = (
                 _get_state_value(state, "llm_input_messages")
             ) or _get_state_value(state, "messages")
-            error_msg = f"Expected input to call_model to have 'llm_input_messages' or 'messages' key, but got {state}"
         else:
             messages = _get_state_value(state, "messages")
-            error_msg = (
-                f"Expected input to call_model to have 'messages' key, but got {state}"
-            )
 
         if messages is None:
-            raise ValueError(error_msg)
+            if pre_model_hook is not None:
+                raise ValueError(
+                    f"Expected input to call_model to have 'llm_input_messages' or 'messages' key, but got {state}"
+                )
+            else:
+                raise ValueError(
+                    f"Expected input to call_model to have 'messages' key, but got {state}"
+                )
 
         _validate_chat_history(messages)
         # we're passing messages under `messages` key, as this is expected by the prompt