Adversarial Consistency and the Uniqueness of the Adversarial Bayes Classifier