Fix: controller: do not check whether watchdog fencing is enabled for the node if stonith-watchdog-timeout is not even configured

Previously controld would unconditionally keep invoking
stonith__watchdog_fencing_enabled_for_node_api() even if
stonith-watchdog-timeout was not configured, which would also keep
worrying users with the log:

pacemaker-controld[...]: notice: Cluster does not have watchdog fencing device