sorts: type recursive_quick_sort for any comparable items

[xiao115255]

What & Why

Part of #15234 — convert sorts/recursive_quick_sort.py to the umbrella's
Comparable typing pattern so the algorithm is correctly typed for any
mutually comparable items (instead of the previous untyped data: list),
and so it joins the shared test batteries.

Concrete changes:

• Rename quick_sort → recursive_quick_sort in
  sorts/recursive_quick_sort.py. The previous name collided with
  sorts/quick_sort.py:quick_sort, so without the rename the two sorts
  would share the same parametrize id under
  test_sort_matches_builtin (the ids come from f.__name__).
• Constrain the items to Comparable with the PEP 695 bounded
  TypeVar [T: Comparable] and list[T] -> list[T], matching the merged
  sibling sorts/quick_sort.py and sorts/bubble_sort.py patterns.
• Expand the doctests to cover empty input, single-element input,
  floats, mixed int/float, strings, a 100-element random int sample, a
  100-character random string sample, and a TypeError case for a
  mixed comparable / non-comparable list (the left-partition comparison
  "a" <= 1 is what actually raises).
• Wire the sort into the shared test batteries in
  tests/test_sorts.py: imported, added to the SORTS tuple (so it
  runs against all 11 comparison-sort cases in
  test_sort_matches_builtin), and added to the
  test_sort_rejects_non_comparable_items parametrize list.

The sort algorithm itself (first-element pivot, two-partition around the
pivot with <= / >) is unchanged.

Part of #15234

How Tested

Local validation on Python 3.14.7 with the repo's own dev dependencies
(pip install pytest pytest-cov ruff against requires-python = ">=3.15"):

• python -m pytest tests/test_sorts.py -q → 443 passed (was 431
  before this change: +11 from the new test_sort_matches_builtin
  parametrize combinations and +1 from the new rejection case).
• python -m pytest --doctest-modules sorts/recursive_quick_sort.py -q
  → 1 passed (the module's own doctests).
• python -m ruff check sorts/recursive_quick_sort.py tests/test_sorts.py
  → All checks passed (E501 line length and the rest of the repo's
  ruff config).
• python -m ruff format --check sorts/recursive_quick_sort.py tests/test_sorts.py → 2 files already formatted.

I also ran the broader sort suite as a sanity check:

• python -m pytest --doctest-modules sorts/ -q → 93 passed (all
  doctests in sorts/).

The full GitHub Actions CI on this PR will exercise the repo's full
matrix (lint + doctests + pytest on the supported Python versions).

AI Disclosure

This PR was prepared with AI assistance (Claude Code, Anthropic). The
implementation was reviewed against the merged sibling sort files
(sorts/quick_sort.py, sorts/bubble_sort.py, sorts/insertion_sort.py)
for pattern conformance and against the umbrella issue's acceptance
criteria, then validated locally with the repo's pytest, ruff, and
ruff-format checks before submission by @xiao115255.

[xiao115255]

CI note: the build check is failing on this branch because of an unrelated, pre-existing flake in web_programming/crypto_price_tracker.py. The doctest makes a live HTTP call to https://api.coingecko.com/api/v3/simple/price?ids=bitcoin&vs_currencies=usd and the runner just received a 403 Forbidden from the upstream (Cloudflare rate-limiting on the shared runner IP). The full log line is:

httpx2.HTTPStatusError: Client error '403 Forbidden' for url 'https://api.coingecko.com/api/v3/simple/price?ids=bitcoin&vs_currencies=usd'
1 failed, 3440 passed in 53.80s

The other four checks on this PR pass: ruff, ty, build_docs (sphinx), and pre-commit.ci. The failing file (web_programming/crypto_price_tracker.py) is unrelated to issue #15234, which is why a fork-PR contributor cannot rerun the failed job without admin rights. A maintainer rerun should turn this green; alternatively, that doctest could be skipped under pytest-run-parallel or mocked, but that is out of scope for #15234.