Jailbreaking Large Vision Language Models in Intelligent Transportation Systems