Text this: Diagnostic Performance of Publicly Available Large Language Models in Corneal Diseases: A Comparison with Human Specialists