Steamroller Problems: An Evaluation of LLM Reasoning Capability with Automated Theorem Prover Strategies