feat(raft): add snapshot and recovery mechanism

[slow2342]

Summary

• Implement leader-initiated snapshots: serialize all user-data CFs (default, mvcc, meta, lease, lease_keys, key_lease) into a single blob, triggered every N applied entries (configurable via snapshot_trigger_log_entries)
• Implement follower snapshot receive and restore: atomic clear + replay CF data in a single WriteBatch, rebuild in-memory indexes
• Stop node on snapshot restore failure to prevent operating on inconsistent state
• 512 MiB snapshot size limit on both create and restore, term() invariant assertion, compact index covers all snapshot-covered entries

Changes

• src/raft/state_machine.rs — create_snapshot() iterates CFs without lock and checks size limit; restore_snapshot() atomic WriteBatch clear+write+rebuild in-memory state
• src/raft/raftrs_store.rs — save_snapshot_data() persists snapshot data and metadata; apply_snapshot() saves data and purges old log entries; snapshot() includes actual snapshot data; term() adds invariant assertion
• src/raft/node.rs — Leader-side snapshot trigger (configurable threshold); follower-side snapshot application; snapshot restore failure stops event loop
• src/config.rs — snapshot_trigger_log_entries config field (default 10000)
• src/auth/role.rs — Remove unused write_perm test helper

Testing

• 79 unit tests pass
• cargo clippy -- -D warnings clean
• cargo fmt -- --check clean

Refs #83