Do Vision-Language Models Understand Visual Persuasiveness?