Support cluster by and distributed by [Spark Branch]

clustered by = distributed by + sort by, so this is related to HIVE-7527. If sort by is in place, the assumption is that we don't need to do anything about distributed by or clustered by. Still, we need to confirm and verify.