feat(cluster): own the VM membership, peer-relation flows, and RAFT recovery (3/7)

[marceloneppel]

Issue

The VM charm's cluster-membership orchestration and RAFT recovery state machine still live charm-side.

Solution

• New ClusterMembershipManager: hosts/units-ip bookkeeping, add_members, add_cluster_member, reconfigure_cluster, update_member_ip, update_endpoint_addresses, departed/raft-removal protocol, the departed-members loop, and the full RAFT recovery state machine (stuck_cluster_check/rejoin/stopped_check/cleanup, raft_reinitialisation with helpers, raft_reconnect).
• PostgreSQLApplication.add/remove_member_ip peer-databag writers.
• events/postgresql.py: VM observers (peer-relation-departed, raft_reconnect) + departed/checks/reconfigure handlers.
• Composition: AbstractPostgreSQLCharm declares the observer custom events and wires the RAFT + membership managers.
• Tests: tests/unit/test_cluster_membership.py (ported from the charm's cases; VM-only, k8s variants skip).

Checklist

• I have added or updated any relevant documentation.
• I have cleaned any remaining cloud resources from my accounts.