Bias and Fairness in Large Language Models: A Survey