Automated Bias, Fairness, and Robustness Testing of Custom Language Models