Improving Text Embeddings with Large Language Models