Regret-Based $(ε,δ)$-optimal Stopping Criteria for Bayesian Optimization | Steady Practice | SteadyPractice