Hamiltonian Monte Carlo Inference of Marginalized Linear Mixed-Effects Models