Update win target to MSVC v145 and enable all possible Release optimizations

[Babouchot]

This aims at enhancing the performance provided by hl.exe using latest compilers and release optimization.
see : https://devblogs.microsoft.com/cppblog/upgrading-c-projects-to-visual-studio-2026/#why-you-should-upgrade-your-c++-build-tools

[Babouchot]

I mark this as draft as though it passes compilation checks on all platforms, the test suite still want to use vs2022 to compile hlc and thus fails because it is not installed by the image. So far I did not have time to troubleshoot where those VS versions are referenced to patch them. You may have an idea about that @yuxiaomao ?

[yuxiaomao]

Makefile template are specified in haxe https://github.com/HaxeFoundation/haxe/blob/development/tests/runci/targets/Hl.hx

And the vs2022 template is from hashlink https://github.com/HaxeFoundation/hashlink/tree/master/other/haxelib/templates/vs2022

[Babouchot]

Thanks for the pointer @yuxiaomao 👍
So I pushed an update to Hl.hx in the haxe test suite and opened a pull request HaxeFoundation/haxe#12894
I suppose it now needs to be merged for the tests to have success here ?

[yuxiaomao]

In order, we need to

• PR hashlink to add a template folder named vs2026 with corresponding template files (see vs2022 folder). ping me if you open that PR so I can merge.
• Merge Haxe side PR
• Then this PR will be run with the correct template

[Babouchot]

I already added the vs2026 template folder in this very PR 🙂

[yuxiaomao]

Oh sorry I didn't check this PR content ^^'
Can you still split this PR, so one PR contains only the addition of vs2026 template ? So I'll be able to merge it + haxe side PR in order (I prefer not merging haxe before hashlink)

[Babouchot]

I just added it here #934

[yuxiaomao]

This PR is ready but I'll need to find some time to test the new build against Shiro's game before merge ^^'

[Babouchot]

No problem I understand. I've been using this version as a custom build for some time now, @trethaller did also for a week or so. But I did not test all games, so there should definitely be some more testing to it.

[Babouchot]

But for information about real world numbers: in the test I did on my machine, this could yield to a ~2ms gain on CPU time in CPU intensive situations.