A Mousetrap: Fooling Large Reasoning Models for Jailbreak with Chain of Iterative Chaos