A performance regression on account of `StringDType`. Affects also at least rank and aggregate, which rely on sort and unique. Keep an eye on https://github.com/numpy/numpy/issues/26510