From Words to Poses: Enhancing Novel Object Pose Estimation with Vision Language Models