Simple LLM Prompting is State-of-the-Art for Robust and Multilingual Dialogue Evaluation