This ensures that vertices are grid-aligned while transforming, instead of just snapping the input values for translate.