[flake8-simplify] Fix incorrect fix for positive maxsplit without separator (SIM905)

[RazerM]

Summary

Resolves #20033

Test Plan

unit tests added to the new split function, existing snapshot test updated.

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[ntBre]

Thanks for jumping on this!

I'm kind of leaning away from defining our own iterator, if we can avoid it.

[ntBre]

I don't think this is safe to do. rfind returns the byte offset of the first character of the match. I tried a case like this:

const fn py_unicode_is_whitespace(ch: char) -> bool {
    matches!(ch, '\u{3000}')
}

fn main() {
    let text = "\u{3000}next";
    let start = text.rfind(py_unicode_is_whitespace).unwrap();
    dbg!(&text[start + 1..]);
}

in the Rust playground, and it panics with a byte index is not a char boundary error. This would be the case for any of the multi-byte whitespace characters in our real py_unicode_is_whitespace function.

[RazerM]

Right, I need to increment by char::len_utf8.

[RazerM]

I'm kind of leaning away from defining our own iterator, if we can avoid it.

Depending how you feel about it we could just remove the fix, since it wasn't spotted in #19851 by me or in review. The iterator is a bit verbose but I think the logic is ok.

[ntBre]

Thanks! I think you're right, I was being overly scared of the iterator. This looks good to me overall, just one idea for simplifying a bit by using (r)split_once. It seemed to work well for me locally, but let me know if I'm still missing something.

[ntBre]

Thank you!

[ntBre]

I took care of the merge conflicts. It looks like Direction mostly just got renamed to Method in #20459.