RFC: prevent mapslices from mutating the original array

[timholy]

This is a broader fix for #17153, making it so that mapslices never modifies the input array. It allocates temporary storage and copies each slice into it before calling the user-function.

@simonster commented (#17153 (comment)) that this would still be breaking compared to the old behavior (which created a new array with A[idx...] on every iteration), but except for "pathalogical functions" (e.g., those whose behavior might depend on the array's pointer address) I confess I fail to see how this is the case.

[timholy]

I should have marked this "RFC" since the proper behavior is of course debatable.

[simonster]

The case I was imagining was pushing slices into an Array and expecting to have n different slices instead of the last slice n times when mapslices returns. I'm not aware of any code that does this, but it's possible.

[simonster]

If this gets merged, we should also change the description of the breaking change to mapslices in NEWS.md, and maybe revert the code changes (but not the tests) from #17154.

[timholy]

Thanks for explaining. I agree this would break such functions, but (as you seem to feel) I don't think we need to bend over backwards to support such uses.

[tkelman]

If this gets merged, we should also change the description of the breaking change to mapslices in NEWS.md, and maybe revert the code changes (but not the tests) from #17154.

should probably do that here before merging

[timholy]

Looks like there are no real objections to this change in strategy.