Consider switching to Veldrid for rendering

Advantages over SlimDX + OpenTK current abstraction:

• one dependency can be used for both Linux and Windows no longer desired
• targets .NET Standard 2.0 (see #1415) since #3702, already done
• Vulkan (see #575)

A proof-of-concept is available on branch veldrid-poc. Checklist for working scenarios on that branch:

• Linux Wayland w/ OpenGL (unimplemented)
• Linux Wayland w/ Vulkan (unimplemented)
• Linux X11 w/ OpenGL (unimplemented)
• Linux X11 w/ Vulkan
• macOS w/ Metal (unimplemented) prerequisites: .NET Core and macOS port, but theoretically supported
• macOS w/ Vulkan (unimplemented) ditto
• Windows w/ D3D11
• Windows w/ OpenGL (unimplemented)
• Windows w/ Vulkan

gonna leave this here for .NET 6 / Avalonia
maybe Ryujinx' PRs 3901 and 3902 can help? it looks like they were using the same Veldrid helpers
edit: gdi those are probably lost forever thanks to Ryujinx being deleted

shader cross-compilation (no idea what this actually is beyond that, found it on a random HN thread): https://github.com/floooh/sokol-tools/blob/master/docs/sokol-shdc.md

Start with the veldrid-poc branch and review the rendering abstraction and its working-scenario checklist. Compare the Linux, Windows, and potential macOS paths, then determine the missing implementation and validation needed before a Veldrid switch could be considered complete; the issue does not define a final completion criterion.