Clustered multiple generalized expected improvement: A novel infill sampling criterion for surrogate models