Fix required immediate slots calculation and panics when slots overflow

[beicause]

Connections
Immediates issue: #8556

Description

1. Fixes Naga immediate_slots_required calculation is incorrect #9724

2. Fixes panic when var<immediate> is larger than 256 bytes / the bitmask overflows.

Testing

For 1, test with CTS webgpu:api,validation,encoding,programmable,pipeline_immediate:required_slots_set:*
For 2, test with CTS webgpu:api,validation,pipeline,immediates:*.

Squash or Rebase?

Squash

Checklist

• I self-reviewed and fully understand this PR.
• WebGPU implementations built with wgpu may be affected behaviorally.
• Validation and feature gates are in place to confine behavioral changes.
• Tests demonstrate the validation and altered logic works.
• CHANGELOG.md entries for the user-facing effects of this change are present.
• The PR is minimal, and doesn't make sense to land as multiple PRs.
• Commits are logically scoped and individually reviewable.
• The PR description has enough context to understand the motivation and solution implemented.

[andyleiserson]

It might be simpler to use a BitVec to store the immediate slots so we don't have to plumb this error everywhere.

Or maybe just say that Naga's "Maximum byte-size of an immediate data variable." is 256 B, and reject such modules in validation.

A third option might be to represent the invalid state internally in ImmediateSlots, so it can can have impl Default and the structs that use it can derive Default. (With a type parameter for error or Infallible, if needed.) But I think one of the other options is probably simpler.

[beicause]

bit_vec::BitVec is heap allocated thus it's not suitable for this.

naga itself doesn't use ImmediateSlots. It only used in wgpu-core.

[andyleiserson]

naga itself doesn't use ImmediateSlots. It only used in wgpu-core

True. I was thinking of oversize immediates as something naga might want to support as valid WGSL for the interface reflection, even if not supported by wgpu or any actual implementation. But there's really no point in that.

But then I feel like it makes even less sense to plumb around ImmediateSlotsOverflowError and break derived Defaults for this case. Or put differently, it's weird for the error ImmediateSlotsOverflowError to only ever be unwrapped, never reported to a user.

What do you think about beicause/wgpu@fix-naga-immediate-slots-required...andyleiserson:jj-push-urxm? (We could also remove the overflow error, or not. I'm okay with the way it's used here, it's the way it's used in wgpu-core that bothers me.)

[andyleiserson]

naga itself doesn't use ImmediateSlots. It only used in wgpu-core

True. I was thinking of oversize immediates as something naga might want to support as valid WGSL for the interface reflection, even if not supported by wgpu or any actual implementation. But there's really no point in that.

But then I feel like it makes even less sense to plumb around ImmediateSlotsOverflowError and break derived Defaults for this case. Or put differently, it's weird for the error ImmediateSlotsOverflowError to only ever be unwrapped, never reported to a user.

What do you think about beicause/wgpu@fix-naga-immediate-slots-required...andyleiserson:jj-push-urxm? (We could also remove the overflow error, or not. I'm okay with the way it's used here, it's the way it's used in wgpu-core that bothers me.)