Text this: DMGD: Train-Free Dataset Distillation with Semantic-Distribution Matching in Diffusion Models