Large Language Models Still Can't Plan (A Benchmark for LLMs on Planning and Reasoning about Change)