Deprecate metadata_request_timeout

When I was testing e4a000f,
I have figure that out there is another option that controls system queries timeout - `Cluster.control_connection_timeout`,
switched to use it instead of `Cluster.metadata_request_timeout`, but did not remove `Cluster.metadata_request_timeout`.

Effectively `Cluster.metadata_request_timeout` is not working, what is working is `Cluster.control_connection_timeout`.
To address that let's mark `Cluster.metadata_request_timeout` it as deprecated in favor of `Cluster.control_connection_timeout`, to be removed in next minor release.

Author: dkropachev

cassandra/cluster.py

@@ -1082,6 +1082,8 @@ def default_retry_policy(self, policy):
 
     metadata_request_timeout = datetime.timedelta(seconds=2)
     """
+    Deprecated: use control_connection_timeout instead.
+
     Timeout for all queries used by driver it self.
     Supported only by Scylla clusters.
     """
@@ -1300,6 +1302,7 @@ def __init__(self,
 
         self.auth_provider = auth_provider
         if metadata_request_timeout is not None:
+            warn("metadata_request_timeout is deprecated: use control_connection_timeout instead", DeprecationWarning)
             self.metadata_request_timeout = metadata_request_timeout
 
         if load_balancing_policy is not None: