A Dual-Purpose Deep Learning Model for Auscultated Lung and Tracheal Sound Analysis Based on Mixed Set Training