LobsDICE: OfflineLearningfromObservationvia StationaryDistributionCorrectionEstimation