Large Language Models for Multi-label Propaganda Detection