Does Reasoning Help LLM Agents Play Dungeons and Dragons? A Prompt Engineering Experiment