Beneath the Surface: Unveiling Harmful Memes with Multimodal Reasoning Distilled from Large Language Models | Synapse