Towards Explainable Harmful Meme Detection through Multimodal Debate between Large Language Models | Synapse