Text this: Dual modality prompt learning for visual question-grounded answering in robotic surgery