I'll believe it when I see it: Images increase misinformation sharing in Vision-Language Models