Pedestrian Attribute Recognition via CLIP based Prompt Vision-Language Fusion