Seol mar théacs é seo: Adversarial robustness guarantees for classification with Gaussian Processes