Enhancing Machine Translation through Advanced In-Context Learning: A Methodological Strategy for GPT-4 Improvement