Implement advance_back_by for str::Chars

[CarloTUM]

Chars already has a specialized advance_by (added in 40cf1f9, counts char start bytes in 32 byte chunks), but DoubleEndedIterator for Chars used the default advance_back_by, so chars().nth_back(n), chars().rev().skip(n) / .rev().nth(n) decoded every char one by one.

This mirrors the forward implementation from the back: count non-continuation bytes in 32 byte chunks via as_rchunks, and if the skipped bytes begin in the middle of a char (whose leading byte therefore wasn't counted), keep its continuation bytes in the iterator. The remainder is skipped char by char.

Benchmarks (./x bench library/coretests --test-args chars_advance, corpora::ru::LARGE, x86_64):

bench	before	after
chars_advance_back_by_0001	0.84–0.88 ns	0.45–0.68 ns
chars_advance_back_by_0010	4.7–4.9 ns	3.4–5.1 ns
chars_advance_back_by_1000	980–1062 ns	106–166 ns
(chars_advance_by_1000, ref)	171 ns	165–177 ns

Tests: test_iterator_advance_back (mirror of the forward test) and test_iterator_advance_back_matches_next_back, which compares advance_back_by(n) / nth_back(n) against repeated next_back() for all n (including past the end) on mixed 1–4 byte strings spanning several chunks, at several start offsets.

r? libs

---

AI disclosure: This change was assisted by GitHub Copilot and Claude. I reviewed and tested it myself.

[rustbot]

Thanks for the pull request, and welcome! The Rust Project has assigned @Darksonn (or someone else) to review your changes, you should hear from them (or someone else) within the next two weeks.

Please see the contribution instructions and our LLM policy for more information.

Why was this reviewer chosen?

The reviewer was selected based on:

• libs expanded to 12 candidates
• Random selection from Darksonn, JohnTitor, Mark-Simulacrum, clarfonthey