TECO: Improving Multimodal Intent Recognition with Text Enhancement through Commonsense Knowledge Extraction