Fix sporadic failures in ClusterAdminRpcServiceTest
The testAddShardReplica is failing sporadically indirectly due to
https://git.opendaylight.org/gerrit/#/c/35097/. The leader's payload
version was -1 so a transaction's BatchedModifications instance was
serialized to the legacy protobuf message type and thus wasn't delivered.
The underlying problem is that RaftActor.updateConfigParams
needs to pass the previous leader's payload version when re-constructing
the Follower instance on RaftPolicy change.
Change-Id: Id6b8da98c145f2e265dddbc1b27384d54e400370
Signed-off-by: Tom Pantelis <tpanteli@brocade.com>