Asymptotically Optimal Planning for Non-myopic Multi-Robot Information Gathering