Text this: Fine-tuning with Hierarchical Prompting for Robust Propaganda Classification Across Annotation Schemas