VALOR-EVAL: Holistic Coverage and Faithfulness Evaluation of Large Vision-Language Models