fix/hack Updater.stop() not working on extreme cases

during test_bootstrap_retries_fail() there is an exception raised (by
design): TelegramError('test')

For a reason I haven't managed to pinpoint the above exception in its
precise timing caused the Updater to be left in a state which is
'self.running == False', but the dispatcher threads already initialized.
This patch identifies this extreme case and makes sure to go over the
stop procedure.

telegram/ext/updater.py

@@ -346,7 +346,7 @@ class Updater(object):
 
         self.job_queue.stop()
         with self.__lock:
-            if self.running:
+            if self.running or dispatcher.ASYNC_THREADS:
                 self.logger.debug('Stopping Updater and Dispatcher...')
 
                 self.running = False