gh-152433: Windows: allow build prim.c for UWP

[thexai]

All changes are trivial:

Remove some run-time dynamic linking and replace some LoadLibrary with LoadPackagedLibrary for UWP.

• Issue: Windows: improve compatibility to build for UWP / Xbox #152433

[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.

[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.

[chris-eibl]

LGTM considering @zooba's #152433 (comment).

cc @colesbury who tracks mimalloc differences from upstream in #113141.

[zooba]

This change looks fine, if we're maintaining the sources separately. If we're trying to track upstream closely, then we'll need them to make the change first. That's @colesbury's call, it seems.

[zware]

Our copy of mimalloc is fairly out of date at this point, and does carry several patches against the equivalent upstream version. Regardless, I think we should be sure that this is done upstream before we merge it ourselves, or it's likely to be lost if and when we update our copy -- ideally, when we do such an update we can do it without carrying any patches forward.

[colesbury]

Yeah, if it's not CPython specific, please also open a PR in upstream mimalloc.

[thexai]

Patch submitted to upstream (microsoft/mimalloc#1340).

Note is a bit different because GetNumaHighestNodeNumber thing is already addressed in current version and only replaced LoadLibrary for better code portability (to very old Windows versions).

I think PR can still be merged here because it doesn't do any harm and in the future it will be overwritten by whatever is upstream.

[zooba]

Not sure it's a major issue, but wouldn't GetModuleHandle be the better option? There's 0% chance we've reached this point without these module being loaded, and as they're not packaged libraries I would expect this to fail without the "it's already loaded" fast-path.

[thexai]

Not seems much different: according the Microsoft doc LoadPackagedLibrary already returns the handle to loaded module in this case:

If the specified module is already loaded in the process, the function returns a handle to the loaded module.

1. The function first searches modules that are already loaded. If the specified module was originally loaded from the package dependency graph of the process, the function returns a handle to the loaded module. If the specified module was not loaded from the package dependency graph of the process, the function returns NULL.

https://learn.microsoft.com/en-us/windows/win32/api/winbase/nf-winbase-loadpackagedlibrary