Improving Retrieval-Augmented Large Language Models via Data Importance Learning