Adversarial Prompt Distillation for Vision-Language Models