Evaluating Large Language Models in Semantic Parsing for Conversational Question Answering over Knowledge Graphs