Decoding hydrometallurgical kinetics of spent lithium-ion batteries via time-series generative adversarial network-driven data augmentation and machine learning to address data scarcity