gh-115882: Add missing #include <Unknwn.h> on Windows

[georgthegreat]

Fix building ctypes module with -DWIN32_LEAN_AND_MEAN defined.

See:
https://learn.microsoft.com/en-us/windows/win32/api/unknwn/nn-unknwn-iunknown

• Issue: Fix building CPython with -DWIN32_LEAN_AND_MEAN on Windows #115882

[cpython-cla-bot]

All commit authors signed the Contributor License Agreement.

[bedevere-app]

Most changes to Python require a NEWS entry. Add one using the blurb_it web app or the blurb command-line tool.

If this change has little impact on Python users, wait for a maintainer to apply the skip news label instead.

[georgthegreat]

Could anyone mark this with skip news label to get the CI green?

[hugovk]

This should have a NEWS entry, please can you add one?

[georgthegreat]

@hugovk, it looks like generating files in Misc/News.d/next is automated somehow.
It there any guide on how do I add an entry atop of the existing PR?

UPD: managed to do this by myself, sorry for disturbing.

[georgthegreat]

NB: this is not the only patch needed to build CPython with this define.

I will have to dig through our patch set to get the remaining ones.

[georgthegreat]

@hugovk, the CI is green, so two questions arise.

1. Can we proceed with merging this PR?
2. Is there a chance for this PR to be backported to 3.12 branch? Updating to 3.13 is a huge amount of work in our case and we would like to remove as many patches as possible prior to doing that.

[hugovk]

1. Can we proceed with merging this PR?

I'll leave it to a Windows expert to review.

2. Is there a chance for this PR to be backported to 3.12 branch? Updating to 3.13 is a huge amount of work in our case and we would like to remove as many patches as possible prior to doing that.

Generally, bug fixes can be backported for all bugfix branches -- see https://devguide.python.org/versions/ -- currently 3.11 and 3.12. Again, I'll leave it to a Windows expert to decide.

[zooba]

This one is fine. But no promises that the rest will be fine - I've definitely fixed issues recently by removing LEAN_AND_MEAN specifications, and I'd rather keep them in than start referencing non-top-level header files directly.

I see no reason to backport, as this doesn't fix a bug (neither us nor Microsoft make any promises about things being buildable with WIN32_LEAN_AND_MEAN). Keep the patches up and you may be able to drop all your own patches for 3.13 onwards, but you'll want to hang onto them for 3.12 and prior.