Preference Elicitation for Step-Wise Explanations in Logic Puzzles