Seol mar théacs é seo: Robust reinforcement learning with Bayesian optimisation and quadrature