Fix a shutdown race in netmgr udp

We need to mark the socket as inactive early (and synchronously)
in the stoplistening process; otherwise we might destroy the
callback argument before we actually stop listening, and call
the callback on bad memory.

lib/isc/netmgr/netmgr.c

@@ -833,10 +833,13 @@ nmsocket_maybe_destroy(isc_nmsocket_t *sock) {
 	if (active_handles == 0 || sock->tcphandle != NULL) {
 		destroy = true;
 	}
-	UNLOCK(&sock->lock);
 
 	if (destroy) {
+		atomic_store(&sock->destroying, true);
+		UNLOCK(&sock->lock);
 		nmsocket_cleanup(sock, true);
+	} else {
+		UNLOCK(&sock->lock);
 	}
 }