FairMonitor: A Four-Stage Automatic Framework for Detecting Stereotypes and Biases in Large Language Models