Evaluating Concept-based Explanations of Language Models: A Study on Faithfulness and Readability