Efficient Conversion of Bayesian Network Learning into Quadratic Unconstrained Binary Optimization