mitigate in-place reducedim alias bug

[Moelf]

fix #39385

it doesn't fix it completely, but I think this fix the most common case.

If user is doing non-trivial viewing and passing part of the original array
into the mutating function, they should check what they are doing?

[LilithHafner]

Documenting that aliasing is prohibited is definitely necessary. Since this PR's creation, that documentation has been added to sum!, but not the other three.

The size equality check feels very much like a bandaid covering up for the lack of a solid aliasing story (#8087). I'm hesitant to merge a solution to the common case that leaves bugs present in more complex edge cases.

[LilithHafner]

Test failures seem relevant.

[Moelf]

I'm hesitant to merge a solution to the common case that leaves bugs present in more complex edge cases.

so we should allow common bug that affects most people to keep being so while we search for a perfect solution that's gonna at least take another year or two to come by?