[krbdev.mit.edu #7753] krb5-1.11: Bug fix: iprop

[Greg Hudson via RT]

There is no locking in ulog_finish_update_slave(), so I can't come up 
with any obvious sequence of events which would result in a delete being 
replayed twice.

So if we do try to replay a delete operation and the principal isn't 
present, I'm not certain whether the right action is to ignore it 
(because the principal is already in the desired state of not existing) 
or force a full dump (because something is clearly wrong with the state 
of the slave DB) as we appear to do now.