DISCRET: Synthesizing Faithful Explanations For Treatment Effect Estimation