Using Visual Language Models to Control Bionic Hands: Assessment of Object Perception and Grasp Inference