Taifead r-phoist: Multi-Output Classification of SMAW Process Parameters from Arc Sound Using MFCC and Deep Audio Embeddings