fix: set _processes_started flag after successful start, not before

If start_processes() raised partway through spawning workers, the flag
was already True, causing a subsequent retry call to be a silent no-op.
Moving the assignment to the end means a failed start remains retryable.

Author: nthmost-orkes

src/conductor/client/automator/task_handler.py

@@ -351,7 +351,6 @@ def stop_processes(self) -> None:
     def start_processes(self) -> None:
         if self._processes_started:
             return
-        self._processes_started = True
         logger.info("Starting worker processes...")
         freeze_support()
         self._monitor_stop_event.clear()
@@ -376,6 +375,7 @@ def start_processes(self) -> None:
             self.stop_processes()
             raise
         self.__start_monitor_thread()
+        self._processes_started = True
         logger.info("Started all processes")
 
     def join_processes(self) -> None: