raft: Apply default replica placement strategy in metadata Raft group

A prior commit implements a ring-based replica placement strategy. This
commit applies that strategy in the statemachine of the metadata Raft
group. This approach ensures all members of the metadata Raft group
yield the same replica groups for the cluster.

Due to the nondeterministic characteristics of a Raft cluster, a node
might have a different review point between its persistent state vs
the actual state of the cluster. For example, consider this scenario.
At the time storage A registers, there is no other storage, hence it
has no replica groups. Right after storage A persists its empty
replica groups, new registration from storage B changes the replica
group list of A in the cluster. Storage A is not aware of this change.

We don't try to address that gap now. Any change of a storage's replica
group is expensive. It might lead to massive data replication or
deletion on a node. So, we'll defer handling replica group changes to
later.