Explaining Patterns in Data with Language Models via Interpretable Autoprompting