if one node is down await schema agreement will never finish

If a node is down it will exist in system.peers but it will never get the new schema because it is down, we need to check our state of the ring and ignore a host if it is down. It would be great if the peers table had gossip node state

No file or test is named. Start by tracing the schema-agreement wait and its ring-state handling, then inspect how system.peers represents unavailable hosts and whether gossip state is available. Done means a down node no longer prevents schema agreement from finishing, with the requested peer-state behavior clarified and verified.

Written by the indexing model from the issue text.