Exploring the Multilingual NLG Evaluation Abilities of LLM-Based Evaluators