On Applying Unit Propagation-Based Lower Bounds in Pseudo-Boolean Optimization