Performance of Large Language Models in ASA Physical Status Classification Using Expert-Validated Synthetic Preoperative Scenarios: A Comparative Study of GPT-4o and Gemini 2.0 | Synapse