gh-152298: Fix LazyImportType.resolve reification

[Kuhai9801]

Make types.LazyImportType.resolve() cache successful reification and replace the original module global only when it still points at the same lazy proxy.

The tests cover cached resolves, stale/rebound/deleted globals, retry after failure, module attribute reification, copied proxies, and reentrant store handling.

Tests:

• git diff --check
• GitHub Actions required checks after squash

• Issue: LazyImportType.resolve() does not replace the module global and can import twice #152298

[python-cla-bot]

All commit authors signed the Contributor License Agreement.

[read-the-docs-community]

Documentation build overview

📚 cpython-previews | 🛠️ Build #33342023 | 📁 Comparing ad578bc against main (1812162)

  🔍 Preview build  

18 files changed · ± 17 modified · - 1 deleted

± Modified

• c-api/type.html
• deprecations/pending-removal-in-3.16.html
• library/asyncio-eventloop.html
• library/asyncio-llapi-index.html
• library/asyncio-runner.html
• library/asyncio.html
• library/curses.html
• library/itertools.html
• library/tkinter.html
• library/tkinter.messagebox.html
• and 7 more...

- Deleted

• library/asyncio-policy.html

[Kuhai9801]

The remaining CI failures appear unrelated to this PR; they fail in Emscripten test_profiling and Windows Tk/IDLE tests, outside the lazy-import changes.

[pablogsal]

I'm struggling really hard to review this or even understand this PR. Has this been ai generated?

[pablogsal]

Ok I am closing this PR please review https://devguide.python.org/getting-started/ai-tools/

[Kuhai9801]

Sorry about that, I accidentally mixed part of gh-152307 into this branch.

I’ll split it back out and keep this focused on gh-152298. Would you prefer a new smaller PR, or should I clean this one up?