Bellman Calibration for Marginalized Importance Weighting in Offline Reinforcement Learning | TickerVault