instantiate dispatcher once per job run

* Instantiating the callback dispatch queue on each job event callback
is expensive. Instead, instantiate it only once. Note, we do not need to
instantiate the callback queue in the iso case so we do not.

awx/main/tasks.py

@@ -1026,9 +1026,8 @@ class BaseTask(object):
             if event_data[self.event_data_key] != 'job_id':
                 event_data.pop('parent_uuid', None)
         should_write_event = False
-        dispatcher = CallbackQueueDispatcher()
         event_data.setdefault(self.event_data_key, self.instance.id)
-        dispatcher.dispatch(event_data)
+        self.dispatcher.dispatch(event_data)
         self.event_ct += 1
 
         '''
@@ -1056,13 +1055,12 @@ class BaseTask(object):
         '''
         Ansible runner callback triggered on finished run
         '''
-        dispatcher = CallbackQueueDispatcher()
         event_data = {
             'event': 'EOF',
             'final_counter': self.event_ct,
         }
         event_data.setdefault(self.event_data_key, self.instance.id)
-        dispatcher.dispatch(event_data)
+        self.dispatcher.dispatch(event_data)
 
     def status_handler(self, status_data, runner_config):
         '''
@@ -1244,6 +1242,7 @@ class BaseTask(object):
                                                            ident=str(self.instance.pk))
                 self.event_ct = len(isolated_manager_instance.handled_events)
             else:
+                self.dispatcher = CallbackQueueDispatcher()
                 res = ansible_runner.interface.run(**params)
                 status = res.status
                 rc = res.rc