Satisficing Exploration in Bandit Optimization