Computational Experiments Meet Large Language Model Based Agents: A Survey and Perspective