Optimal Nonlinear Online Learning under Sequential Price Competition via s-Concavity