3D Vision Language Models (VLMs) for Robotic Manipulation: Opportunities and Challenges