Large Language Models Show Comparable Response Performance but Vary in Readability Regarding Patient Questions on Hip Arthroscopy