The Knowledge of Large Language Models Regarding Response Evaluation Criteria in Solid Tumors: A Comparative Study with Prompt Effect