Causal inference in high dimensions: A marriage between Bayesian modeling and good frequentist properties