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