Optimal Strategies in Perfect-Information Stochastic Games with Tail Winning Conditions