Time-consistent approximations of risk-averse multistage stochastic optimization problems