Maximizing Kinetic Information Gain of Markov State Models for Optimal Design of Spectroscopy Experiments.