Distributed learning for stochastic generalized Nash equilibrium problems