Eluder-based Regret for Stochastic Contextual MDPs