Save cluster state to a file

[andrewjstone]

Currently cluster state is only maintained in memory. We should save it to a file so an admin doesn't need to rejoin all the peers every time on startup.

[andrewjstone]

This is really only a problem if all nodes go down. Otherwise a single node in the quiescent cluster state will continuously try to reconnect to all nodes it knows about. Nonetheless, it is important not to have to reconfigure the cluster even in diskless log mode.