Shutdown corosync after a fatal error

If pacemaker shuts down due to being fenced by a non-power (eg fabric)
fence agent then it should also take down corosync so that full cluster
service on that node is lost, rather than just resource management.

https://bugzilla.redhat.com/show_bug.cgi?id=1448221

Signed-off-by: Christine Caulfield <ccaulfie@redhat.com>