Towards a Framework for Evaluating Explanations in Automated Fact Verification