Make Imagination Clearer! Stable Diffusion-based Visual Imagination for Multimodal Machine Translation