Learning Large-Scale Competitive Team Behaviors with Mean-Field Interactions