Log-Runtimes of linear regression (sklearn) v.s. CATE for exhaustive subgroup search