provider/aws: fix aws_elasticache_replication_group for Redis in clus…

…ter mode (#9601)

This is a fix for issue hashicorp/terraform#9596.

Changes:
 - Adds new output attribute `configuration_endpoint_address`. Only
   used in Redis when in cluster mode.
 - Read the `snapshot_window` and `snapshot_retention_limit` from
   the
   replication group description instead of the cache cluster
   description.
 - Adds acceptance test and modifies an existing acceptance test to
   make sure that everything is still good in non-cluster mode
 - Updates docs to describe new output attribute

r/elasticache_replication_group.html.markdown

@@ -67,13 +67,14 @@ Please note that setting a `snapshot_retention_limit` is not supported on cache.
 
 The following attributes are exported:
 
-* `id` - The ID of the ElastiCache Replication Group
-* `primary_endpoint_address` - The address of the endpoint for the primary node in the replication group
+* `id` - The ID of the ElastiCache Replication Group.
+* `primary_endpoint_address` - The address of the endpoint for the primary node in the replication group. If Redis, only present when cluster mode is disabled.
+* `configuration_endpoint_address` - (Redis only) The address of the replication group configuration endpoint when cluster mode is enabled.
 
 ## Import
 
 ElastiCache Replication Groups can be imported using the `replication_group_id`, e.g.
 
 ```
 $ terraform import aws_elasticache_replication_group.my_replication_group replication-group-1
-```
+```