Modernize to new wgpu features

[Vipitis]

Finding motivation to work on this project some more.

Couple of improvements from upstream advancements:

• use new context
• use wgpu.structs
• use auto layout: currently breaks two examples, likely due to how we attach stubs for unused inputs. Perhaps it's not for our usecase: https://www.w3.org/TR/webgpu/#default-pipeline-layout -> the built auto layout is smarter than our input detection, but we can't access it's entries -.- so I will revert this commit
• use push constants postponed to a different PR, as there is some native only features, but also idl features: https://gpuweb.github.io/gpuweb/wgsl/#pipeline-overridable hopefully included with wgpu release by end of year: Immediates tracking issue gfx-rs/wgpu#8556
• more typing fixes
• Bump CI

[Vipitis]

CI failure is related to #57 and the changes to the Shadertoy.com website like cloudflare ddos protection. The easiest solution would be to just package the texture (and other media) or provide an alternative source. However I am not sure what the individual copyright is for these files. We can also hope that embedding, API and perhaps media becomes available in the near future again.

Generally there should be some fallback for offline use. Maybe we could include a cache filling command for the cli, but feels convoluted.

In general I also want to make using Shadertoy media more user friendly by letting users create channels using the name. For example all three variants could be made to work:

shader = Shadertoy(my_code, inputs=["Bayer", ShadertoyChannel("Font 1"), ShadertoyChannelTexture("Lichen")]

[Korijn]

Happy to see the progress made here! (I've been following along with every commit) Nice work. :)