Test corrade_add_resource() with Unicode filenames

[mosra]

While everything works great on Linux and macOS, this causes CMake with Ninja on Windows to go into an infinite loop. As far as I know, this didn't happen with MinGW makefiles, so shouldn't be an issue with my CMake dependency handling code. It also doesn't seem to happen with VS project files, so probably just related to Ninja on Windows (and/or CMake Ninja support).

ninja: error: manifest 'build.ninja' still dirty after 100 tries

Putting this here so it doesn't get forgotten.

Things left to do:

• Look into this again, maybe it's finally fixed five years later? not really, what am I doing wrong?
• Ninja apparently doesn't properly convert to UTF-16 and still uses the ANSI APIs, sigh -- it was attempted in Ninja UTF8 conversion: Using native windows API for conversion. ninja-build/ninja#1671 but that was closed, hopefully it'd be done along with long path support in Add long path support for Windows [Approach #2] ninja-build/ninja#2056 ?
  • But at least Ninja 1.11 opts in to Windows 10 UTF-8 support, so it's just older Windows versions left (Use UTF-8 on Windows 10 Version 1903, fix #1195 ninja-build/ninja#1915) -- download that
  • And CMake 3.21+ uses UTF-8 when writing Ninja files (https://gitlab.kitware.com/cmake/cmake/-/merge_requests/5860) -- MSVC 2019 and 2022 AppVeyor jobs have 3.24
    • Once it starts working there, disable the test on older images?
  • But Appveyor uses Windows Server 2019 which might not include the required UTF-8 patches from Windows 10?
• This should also be a regression test case for 90f357a -- try reverting it and verify the build indeed fails with it missing