On the Optimality of the Median-of-Means Estimator under Adversarial Contamination