Certification of Speaker Recognition Models to Additive Perturbations