Approachability in repeated games: Computational aspects and a Stackelberg variant