Don't destroy a non-closed socket, wait for all the callbacks.

Don't destroy a non-closed socket, wait for all the callbacks.

We erroneously tried to destroy a socket after issuing
isc__nm_tcp{,dns}_close. Under some (race) circumstances we could get
nm_socket_cleanup to be called twice for the same socket, causing an
access to a dead memory.

(cherry picked from commit 233f134a4ff806187925a106270a1f3794d5a1d5)