Predicting positive airway pressure appointment attendance using machine learning and deep learning: the role of clinical and oximetric indicators