[DocDB] Longer recovery from Master failures after recent changes

[amitanandaiyer]

Description

Master rpcs should retry if the master-leader is unreachable.
This is currently done based on whether the error is a NetworkError/RemoteError

With https://phabricator.dev.yugabyte.com/D16073 / https://github.com/yugabyte/yugabyte-db/commit/ 1b8f992

some NetworkErrors are now being recategorized as HostUnreachable.

Thus, if a master node becomes unavailable, we should handle the HostUnreachable error similar to
what we are doing for NetworkError to the master.

Also add more details to the error message in HostUnreachable