fix(ssubscribe): [cherrypick from v4] Force slots refresh on MOVED error when using ssubscribe

[htemelski-oss]

When issuing a ssubscribe command, the Cluster client creates a "regular" Redis client to establish a connection to the correct shard. This doesn't work if there's a failover since the client doesn't know how to correctly handle a MOVED response. This PR handles such events by propagating the ssubscribe MOVED error from the DataHandler, through the ClusterSubscriber and the ClusterSubscriberGroup to the RedisCluster client and forces slots refresh.

Direct port from #1987