Two-step "win-stay, lose-shift" and learning to cooperate in the repeated Prisoner's Dilemma