Large Language Models for Propaganda Detection