Implement Send/Sync non-structurally for wgpu wrapper types

[SkiFire13]

Objective

• Improve compile times in bevy_render
• Helps with bevy_render 0.19.1 doesn't compile on current Rust nightly due to -Znext-solver enabled by default #25511
• This is an alternative to Switch WgpuWrapper to a macro and have it compute Send/Sync eagerly #25512 with roughly the same compilation time improvements

Solution

• Use non_structural_derive to short-circuit the Send/Sync checking
• Compared to Switch WgpuWrapper to a macro and have it compute Send/Sync eagerly #25512 this is not a breaking change, however it is less robust: we might unintentionally add a new wgpu wrapper type without the non-structural annotation, and this would regress without visible errors. It also requires an additional dependency (or a bunch of boilerplate to do it ourselves)
  • This could be used for a non-breaking 0.19.2 update, while picking Switch WgpuWrapper to a macro and have it compute Send/Sync eagerly #25512 as a more robust solution for 0.20

Testing

Not needed

Click to view showcase

println!("My super cool code.");

[beicause]

If this is for 0.19, should it target a branch for 0.19 rather than the main branch?

[SkiFire13]

Yes, this would be for an eventual 0.19.2. Should I target 0.19.1 or some other branch?