Hierarchical Bayesian bootstrap for heterogeneous treatment effect estimation