VAP-Diffusion: Enriching Descriptions with MLLMs for Enhanced Medical Image Generation