Automatic Fairness Testing of Neural Classifiers through Adversarial Sampling