Implementing Rational Choice Functions with LLMs and Measuring their Alignment with User Preferences