Describir: Getting More out of Large Language Models for Proofs