Deep lifted decision rules for two-stage adaptive optimization problems