fix: use BoundEvent.put() in engine error handlers to avoid unawaited coroutine warnings

The error handlers _on_error_handler() and _send_error_execution() used
sm.send() which triggers _processing_loop() — creating an unawaited
coroutine in AsyncEngine. Since these are always called within an active
macrostep, BoundEvent.put() is sufficient to enqueue the error.execution
event on the internal queue without the redundant processing loop call.

Author: fgmacedo

statemachine/engines/base.py

@@ -159,7 +159,7 @@ def handler(error: Exception) -> None:
             # new error.execution is a separate event that may trigger a different
             # transition (see W3C test 152).  The infinite-loop guard lives at the
             # *microstep* level (in ``_send_error_execution``), not here.
-            self.sm.send(_ERROR_EXECUTION, error=error, internal=True)
+            BoundEvent(_ERROR_EXECUTION, internal=True, _sm=self.sm).put(error=error)
 
         return handler
 
@@ -188,7 +188,7 @@ def _send_error_execution(self, error: Exception, trigger_data: TriggerData):
         if trigger_data.event and str(trigger_data.event) == _ERROR_EXECUTION:
             logger.warning("Error while processing error.execution, ignoring: %s", error)
             return
-        self.sm.send(_ERROR_EXECUTION, error=error, internal=True)
+        BoundEvent(_ERROR_EXECUTION, internal=True, _sm=self.sm).put(error=error)
 
     def start(self, **kwargs):
         if self.sm.current_state_value is not None: