Get rid of the tricks to avoid overflow in the AbstractUnitRange iterator

[haampie]

Thanks to the new iterator protocol we can now iterate over a unit range in the number type of its elements, without having to worry about overflow :). Fixes #26586.

Previously iterating over say typemin(Int8) : typemax(Int8) was done by using an Int internally which was converted to an Int8 each iteration. Now we can just iterate with Int8.

Something similar can probably be done for StepRange, but that would be another pr.

[KristofferC]

Just to check, there already exists tests for e.g. iterating

typemin(Int8) : typemax(Int8)

?

[haampie]

Yes: https://github.com/JuliaLang/julia/blob/master/test/ranges.jl#L421-L426 :)

[haampie]

Test failures seem unrelated (libgit, distributed), ranges tests are green everywhere :)

[StefanKarpinski]

Who would be good to review this?

[Keno]

Yep, that's how this is supposed to work :)

[Keno]

Let's run nanosoldier though, since LLVM is sensitive to how ranges are lowered:
@nanosoldier runbenchmarks(ALL, vs = ":master")

[nanosoldier]

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @ararslan