`<[T]>::sort_by_index` and `<[T]>::sort_by_key_and_index`

It seems to be the issue with short-circuiting OR, as it creates too many bloat for optimizer. But there, I believe, llvm should be able to remove bounds checks for slice. But yeah, I don't know if assert of equal length for each array before the sort would still hold inside the closure.

Could you elaborate a little bit more on this api? I can't quite grasp what is it. And do I understand correctly, that it will require allocation at the end?