Normative Evaluation of Large Language Models with Everyday Moral Dilemmas