Optimal Order Simple Regret for Gaussian Process Bandits