Hackish bug fix for hanging connections on large queries

If the query is large, psycopg2 sets async status to WRITE and requires
additional calls to poll() to flush the query to postgresql.  Momoko was
telling tornado to only listen for READ, so poll() was never called and
the query would hang

momoko/pools.py

@@ -318,8 +318,8 @@ def cursor(self, function, function_args, callback, cursor_kwargs={}):
         getattr(cursor, function)(*function_args)
         self._callbacks = [partial(callback, cursor)]
 
-        # Connection state should be 1 (write)
-        self._ioloop.update_handler(self._fileno, IOLoop.READ)
+        # Fire callback immediately To correctly set the update handler
+        self._io_callback(None, None)
 
     def _io_callback(self, fd, events):
         try: