LLM+AL: Bridging Large Language Models and Action Languages for Complex Reasoning about Actions