Bayesian Optimization under Heavy-tailed Payoffs