SortUtils: fix termination condition & enforce unsigned value

[sciecode]

Fixes incorrect termination condition during insertion-sort block, exposed by #28401
Informs & enforces unsigned 32b integer sort.

---

There was a pretty nasty bug in the termination condition of insertion sort, fortunately, it didn't cause wrong results under expected input conditions (uint32). It only became apparent when used with signed values. Nonetheless, it was fixed and unsigned condition is enforced even if input values don't match expected range.

[gkjohnson]

Thanks for tracking this down! This is a good change, I think, so we don't remove items from the original list. But negative numbers will still result in an incorrect sort, right? Which is okay I just want to make sure I understand.

[sciecode]

But negative numbers will still result in an incorrect sort, right?

That is correct, negative numbers will still be sorted by their 2-complements representation. AKA if you simply bitcast a signed to unsigned in languages like c/c++. Unfortunately radix-sort can't, own its own, solve the problem of signed sorting. It is usually controlled by adjusting the input-keys, like we do in the batch example.

So essentially, this PR is just to make sure that if negative values / overflown 32-bit are inputted, we just don't fail catastrophically as it did previously.