Evaluating the Ability of Large Language Models to Identify Adherence to CONSORT Reporting Guidelines in Randomized Controlled Trials: A Methodological Evaluation Study