Seol mar théacs é seo: Using constraint preconditioners with regularized saddle-point problems