raft: Add AddReplica RPC for operator-driven cluster membership

For #6908

Operators currently lack a safe, high-level API to add replicas to Raft partitions during cluster scaling or failure recovery. The only option is using internal AddNode() and AddLearner() methods that require deep understanding of Raft internals, manual leadership verification, and routing table coordination. This forces operators to rely on complex manual intervention, increasing operational risk and making cluster management error-prone.

This MR introduces the AddReplica RPC that provides a safe, operator- friendly interface for adding replicas to Raft clusters. The RPC abstracts Raft complexity by automatically handling leader verification, partition discovery across storages, and replica bootstrapping with rollback support if the operation fails.