Make sure system.peers and peers_v2 match cluster metadata

[minal-kyada]

Description:
system.peers and system.peers_v2 can drift out of sync with ClusterMetadata, causing clients who use older C* version and tools that read these legacy tables to observe incorrect cluster topology.

Adds SystemPeersValidator which reconciles both peer tables against ClusterMetadata on startup- removing stale entries for nodes no longer in the cluster and repairing missing entries for joined nodes. Also exposes this as a JMX operation via StorageServiceMBean so operators can trigger it on demand without restarting.

Quick diagram of what the patch is about:

patch by @minal-kyada; reviewed by @ifesdjeen @krummas for CASSANDRA-21187

[ifesdjeen]

we should probably throw if it's null;

also, a small nit: I wouldn't nest entire method within if's brackets, i would early-exit on unmatched condition, and unindent the rest of the method.

[ifesdjeen]

Don't we want to validate all fields of the table? In updateLegacyPeerTable, we're updating more than just these two fields:

            QueryProcessor.executeInternal(String.format(peers_v2_query, SYSTEM_KEYSPACE_NAME, PEERS_V2),
                                           addresses.broadcastAddress.getAddress(), addresses.broadcastAddress.getPort(),
                                           addresses.broadcastAddress.getAddress(), addresses.broadcastAddress.getPort(),
                                           addresses.nativeAddress.getAddress(), addresses.nativeAddress.getPort(),
                                           location.datacenter, location.rack,
                                           nodeId.toUUID(),
                                           next.directory.version(nodeId).cassandraVersion.toString(),
                                           next.schema.getVersion(),
                                           tokens);