dlm_controld: use new quorum api to detect missed failures
    
    If a node failed and then quickly rejoined before corosync
    detected the original failure, the quorum callback from
    corosync would not include an indication that the node had
    ever failed.  This meant there would be no removal and
    readdition of the node's configfs settings in the kernel,
    and the failed network connections may never be cleared.
    
    The new quorum nodelist callback does include an indication
    of a node that has left and rejoined.  This patch keeps track
    of nodeids that have left and rejoined through that callback
    between sequential occurances of the standard quorum callback.
    
    When the standard quorum callback is removing and adding
    configfs settings for nodes that have been removed or added,
    it will also now look at the list of nodes that have left
    and rejoined, and do configfs removal and addition for those.