sanlock: fix host_id reads from paxos_acquire
    
    the threads doing paxos_acquire were copying the host_id_disk,
    including fd's, and reading the host_id disk using those fd's,
    (in use by the host_id thread) instead of opening the host_id
    disk themselves to read from.  This probably explains why the
    host_id leader reads sometimes returned the wrong data.
    
    Also add -R and -Q options to enable renewal debugging and
    quiet acquire failure messages.