Alternating proximal-gradient steps for (stochastic) nonconvex-concave minimax problems