gh-122549: Allow to invalidate platform's cached results

[picnixz]

Using a cache invalidation approach is probably better than querying the new results every time. At least, it will not cause any breaking change.

• Issue: Add platform.invalidate_caches for invalidating cached results #122549

---

📚 Documentation preview 📚: https://cpython-previews--122547.org.readthedocs.build/

[kashav]

The current implementation doesn’t seem useful to me.

[vstinner]

LGTM. The function looks useful when the hostname changes, tests look complete, and the documentation is good enough.

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @picnixz for commit 42668a3 🤖

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

[picnixz]

I just want to be sure that build bots won't be unhappy with this change for whatever reason (we'll get the answer in roughly 8 hours)

[picnixz]

The USAN failures are known #111178, the refleaks on Fedora are known as well #122136.

[picnixz]

I should have used a versionadded:: next here but at that time, it didn't exist yet. Not sure if this warrants an additional commit though. WDYT @vstinner

[vstinner]

next will be replaced with 3.14 anyway, there is no need to replace it.

[picnixz]

Yea but sometimes users are confused that the feature is present in the docs but not in the current release. That was the purpose of next