Low: pacemakerd: handle shutdown requests better

Previously, "pacemakerd --shutdown" would loop forever sending CRM_OP_QUIT
IPC requests to the existing pacemakerd instance, until IPC went away
(which will never happen if the requests continually fail).

Now, it checks for (and bails on) errors, and uses a 30-minute timeout
waiting for IPC to go away after a successful request.