Exploring In-Context Learning for Frame-Semantic Parsing