Update TokenAwarePolicy.make_query_plan to schedule to replicas first

End result should be like this:
1. Replicas from the same RACK (if rack is specified)
2. Replicas from the same DC, but Remote RACK (if rack is specified)
3. Replicas from the same DC (if rack is not specified)
3. Replicas from the remote DC
5. Non-replicas in the same order

Author: dkropachev

cassandra/policies.py

@@ -530,14 +530,21 @@ def make_query_plan(self, working_keyspace=None, query=None):
         if self.shuffle_replicas:
             shuffle(replicas)
 
-        for replica in replicas:
-            if replica.is_up and child.distance(replica) in [HostDistance.LOCAL, HostDistance.LOCAL_RACK]:
-                yield replica
+        def yield_in_order(hosts):
+            for replica in hosts:
+                if replica.is_up and child.distance(replica) == HostDistance.LOCAL_RACK:
+                    yield replica
 
-        for host in child.make_query_plan(keyspace, query):
-            # skip if we've already listed this host
-            if host not in replicas or child.distance(host) == HostDistance.REMOTE:
-                yield host
+            for replica in hosts:
+                if replica.is_up and child.distance(replica) == HostDistance.LOCAL:
+                    yield replica
+
+            for replica in hosts:
+                if replica.is_up and child.distance(replica) == HostDistance.REMOTE:
+                    yield replica
+
+        yield from yield_in_order(replicas)
+        yield from yield_in_order([host for host in child.make_query_plan(keyspace, query) if host not in replicas])
 
     def on_up(self, *args, **kwargs):
         return self._child_policy.on_up(*args, **kwargs)