Stopping Criteria for Value Iteration on Stochastic Games with Quantitative Objectives