Development of an interpretable machine learning model for predicting prolonged mechanical ventilation in adults after cardiac valve surgery