Enhancing Quantitative Reasoning Skills of Large Language Models through Dimension Perception