PAVLM: Advancing Point Cloud based Affordance Understanding Via Vision-Language Model