Decompiling the synergy: An empirical study of human–LLM teaming in software reverse engineering