KOCO-BENCH: Can Large Language Models Leverage Domain Knowledge in Software Development?