Fix: crmd: Resend the shutdown request if the DC forgets

As seen in:

https://bugzilla.redhat.com/show_bug.cgi?id=1310486

Scenario needs very poor timing and some bad luck:

1. Start a node wait for it to become the DC
2. Start a second node
3. Tell the second node to stop while it is in the process of

negotiating with the DC.

Specifically just after do_cl_join_finalize_respond() is called on

the second node.

4. Cross your fingers that somehow the shutdown=0 update makes it to

attrd _after_ the DC sets shutdown=${large int}

Under these conditions, the request to shut down will be lost and the DC
will feel free to start services on the second node.