Reprise of #7875, which was reverted in #7905.

This tightens up the validation of binding sizes. I have added a helper on the wgpu_core Buffer type to resolve implicit remainder-of-buffer sizes and validate the binding offset and size against the size of the buffer. This helper is now the preferred way of constructing a hal::BufferBinding. Direct construction is no longer allowed in wgpu_core. There is an unsafe new_unchecked helper for cases where direct construction is necessary.

This fixes a small but critical typo in the previous change: 14a5abd

It also stays even closer to the original behavior for zero-size bindings than the previous change. When we validate a binding with implicit remainder-of-buffer size, we preserve the information that it is sized as such, and resolve it again in the backends.

I've divided the changes into meaningful independent commits. Since #7875 was reverted, the PR diff is against the state prior to those changes. The diff against the state after #7875 may also be informative: https://github.com/gfx-rs/wgpu/compare/e5b03ffa1defc0b760a8c0bd8c54aa9b98c78002...andyleiserson:wgpu:buffer-size-fixes?expand=1.

Testing
Enables some CTS tests that were broken by the previous change.

The following CTS tests are still failing, but are also relevant:

webgpu:api,validation,encoding,cmds,render,draw:vertex_buffer_OOB:*
webgpu:api,validation,encoding,cmds,render,draw:buffer_binding_overlap:*

Squash or Rebase? Undecided

Checklist

• Run cargo fmt.
• Run taplo format.
• Run cargo clippy --tests. If applicable, add:
  • --target wasm32-unknown-unknown
• Run cargo xtask test to run tests.
• If this contains user-facing changes, add a CHANGELOG.md entry.