De-AntiFake: Rethinking the Protective Perturbations Against Voice Cloning Attacks