Update wgpu to 0.19, glyphon to 0.5, softbuffer to 0.4, window-clipboard to 0.4, and raw-window-handle to 0.6

[ids1024]

Based on #2169 and #2190. This updates wgpu to the latest trunk (the last release doesn't use raw-window-handle 0.6).

This can't be merged until those two PRs are, and when a wgpu release is made. I just wanted to test Iced with newer Wgpu changes. And to see how the API changes here.

This should make the API here safe, while RawWindowHandle/RawDisplayHandle can be unsafe to use, since they can contain pointers and don't constrain the lifetime.

Problems:

• To create a softbuffer Context and Surface from W, we need W: Clone. So this adds an Arc around winit::window::Window.
• wgpu::WasmNotSendSync bounds are a bit awkward.
• It should be possible for a Compositor to handle non-'static W. Need to try getting that to work.
• window_clipboard will need to be updated for raw-window-handle 0.6

If the way softbuffer and wgpu use WindowHandle/DisplayHandle is problematic for iced, perhaps changes could be made there.

Closes #2190.

[ids1024]

Wgpu 0.19 has just been released.

But this depends on grovesNL/glyphon#80, and an update to window_clipboard.

[ids1024]

I think this should basically be ready pending glyphon and hecrj/window_clipboard#24. If everything is working properly.

[hecrj]

Looks great! Thanks! 🥳

I have simplified some stuff here and there. Let me know what you think and we can merge.

[ids1024]

Those changes seem reasonable. Out-of-order ages won't currently happen with softbuffer, but might in the future if it makes sense for any backend, so that seems reasonable. Using the max age that's been seen seems like a better idea than what I had. It means not having damage tracking for the first couple frames (which perhaps is why I didn't think of it), but that's reasonable.

[hecrj]

@ids1024 Awesome!

Currently blocked by gfx-rs/wgpu#5088.

[ids1024]

We could probably detect if the dx12 backend is being used an then set desired_maximum_frame_latency to 3 (which perhaps is what a fix in wgpu is going to do?). Though if they release a fix fairly soon that isn't really necessary.

[dtzxporter]

We could probably detect if the dx12 backend is being used an then set desired_maximum_frame_latency to 3 (which perhaps is what a fix in wgpu is going to do?). Though if they release a fix fairly soon that isn't really necessary.

This seems a bit hacky, I've spoken with the wgpu team and they intend to release a patch release for this fix. I already have an approved PR waiting to be merged that fixes the issue.

[thunderstorm010]

gfx-rs/wgpu#5088 is fixed by gfx-rs/wgpu#5091!

[hecrj]

Let's merge then!