Cognitive Biases in Large Language Models: A Survey and Mitigation Experiments