[spv-in] Improve detection of ImageClass

[cwfitzgerald]

Small thing I found that if you write Texture2D<uint> in slang, it will add a format of R32Uint to the image, which tripped up our detection. This makes the behavior better. If anyone runs into this and doesn't have this patch, add [[vk::image_format("unknown")]] and it will generate the spirv we expect.

[Copilot]

Pull request overview

This PR improves the SPIR-V input frontend's detection of ImageClass to correctly handle cases where textures have explicit formats specified (e.g., Texture2D<uint> in Slang generating R32Uint format). The fix refines the logic for determining whether an image is a storage texture, sampled texture, or depth texture based on the combination of the is_sampled flag and explicit format specification.

Changes:

• Enhanced ImageClass detection logic to properly handle unknown sampling information with explicit formats
• Improved error messages to distinguish between different multisampling-related validation failures
• Added validation to prevent depth images from being storage textures

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
naga/src/front/spv/mod.rs	Updated ImageClass detection logic to infer storage class from explicit format when sampling info is unknown
naga/src/front/spv/error.rs	Renamed error variant and added new error for depth+storage combination
naga/src/valid/expression.rs	Split InvalidImageOtherIndex into two specific errors and added debug statement

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Debug print statement (std::dbg!) should be removed before merging. This appears to be leftover debugging code that will pollute stderr output in production.

Suggested change

std::dbg!(e, class);

[Wumpf]

nice, great errors. but leftover dbg needs to go. Pre-approving

[Wumpf]

leftover dbg