Performance evaluation of large language models with chain-of-thought reasoning ability in clinical laboratory case interpretation.