Make compute pass end consume the pass

[Wumpf]

Connections

• Follow-up & based on Remove lifetime constraints from wgpu::ComputePass methods #5570
  • ⚠️ Will rebase once #5570lands as this contains all the changes again.
• Puzzle piece, or rather "convenience refactor" on the way to Lifetimes on RenderPass make it difficult to use.  #1453

Part of a series towards lifetime removal on compute passes:

• Remove exposed C symbols from renderpass/computepass recording #5409
• Separate out ComputeCommand id->arc resolve (a step towards no lifetimes on wgpu::ComputePass) #5432
• Remove lifetime constraints from wgpu::ComputePass methods #5570
• ➡️ Make compute pass end consume the pass #5575
• Remove lifetime dependency of ComputePass to its parent command encoder #5620
• Fix QuerySet ownership of ComputePass #5671

Description

• avoid copies on compute pass end:
  Some work in Separate out ComputeCommand id->arc resolve (a step towards no lifetimes on wgpu::ComputePass) #5432 highlighted that a compute pass (render pass as well, but let's ignore this for now) isn't fully consumed during pass ending which means we have to do a couple of copies despite the fact that a pass can not be consumed twice.
• error on ending compute pass twice:
  Any operation on an ended compute pass now causes an validation error (as it should per spec!). As expected this allows to remove a few copies (including needless reference count increment + decrement!). Naturally, this means now that all operations on the compute pass are failable
• rename run to end, make end operation part of compute pass:
  Previously, ending a pass was seen as an operation on the command encoder across the api/crate layers. Now it is part of the render/compute pass in accordance with the WebGPU spec. This is relatively minor but also comes with an error check.
• improve error handling:
  Compute pass errors make now use of its label

Testing
Existing tests still pass, should be enough for this one.

Checklist

• Run cargo fmt.
• Run cargo clippy. If applicable, add:
  • --target wasm32-unknown-unknown
  • --target wasm32-unknown-emscripten
• Run cargo xtask test to run tests.
• Add change to CHANGELOG.md. See simple instructions inside file.

[ErichDonGubler]

praise: These changes make sense to me. Glad to see them!

nitpick: I've taken the liberty of modifying history a bit here, so review is easier and that we might reasonably consider rebasing instead of squashing. I've made individual conversations where it made sense. For general changes, I've folded the following commits into parents: d781e30, e07ba12, 6315094

[ErichDonGubler]

question(non-blocking): Why add #[inline] here? Was it necessary to maintain some performance characteristic alongside these other changes?

[ErichDonGubler]

I'll leave this PR un-merged for now, not because I'm not confident it'll land as-is, but because I want to give @Wumpf a chance to look over changes I've made to this PR's history before they enter mainline.