cluster: default metadata_request_timeout to control_connection_timeout

metadata_request_timeout and control_connection_timeout are going hand
by hand, control_connection_timeout is client side timeout, while
metadata_request_timeout is server side.
In general case they always have to be eual.

Author: dkropachev

cassandra/cluster.py

@@ -1083,10 +1083,11 @@ def default_retry_policy(self, policy):
     used for columns in this cluster.
     """
 
-    metadata_request_timeout = datetime.timedelta(seconds=2)
+    metadata_request_timeout = None
     """
     Timeout for all queries used by driver it self.
     Supported only by Scylla clusters.
+    Defaults to control_connection_timeout.
     """
 
     @property
@@ -1303,8 +1304,7 @@ def __init__(self,
         self.no_compact = no_compact
 
         self.auth_provider = auth_provider
-        if metadata_request_timeout is not None:
-            self.metadata_request_timeout = metadata_request_timeout
+        self.metadata_request_timeout = self.control_connection_timeout if metadata_request_timeout is None else metadata_request_timeout
 
         if load_balancing_policy is not None:
             if isinstance(load_balancing_policy, type):