DAM-VLA: A Dynamic Action Model-Based Vision-Language-Action Framework for Robot Manipulation