feat(graphics): adopt WGSL uniform_buffer_standard_layout for tighter uniform-buffer packing

Background

Chrome 144 added the WGSL uniform_buffer_standard_layout language feature. When enabled via requires uniform_buffer_standard_layout;, uniform buffers may use the same memory layout constraints as storage buffers — in particular, the 16-byte alignment requirement on array elements is lifted. This matches GLSL's std430 semantics.

Why this matters

The engine packs uniform data into uniform buffers automatically (via the uniform buffer / dynamic buffer paths). Today, arrays like array<f32, N> or array<vec3<f32>, N> get padded out to 16-byte stride, wasting bandwidth and complicating layouts shared between uniform and storage buffers.

With this feature, the engine could:

• Pack array<f32, N> and array<vec3<f32>, N> at natural stride
• Share data structure layouts between uniform and storage buffers without padding hacks
• Reduce uniform buffer bandwidth for the affected types

Why this is more work than a typical WGSL feature cap

Unlike the other WGSL feature caps (which just emit a requires directive and let shaders opt-in conditionally), the engine's uniform buffer packing code in JS (e.g. UniformBuffer, UniformBufferFormat, the dynamic buffer system) currently encodes the std140-style alignment rules. To benefit from uniform_buffer_standard_layout, the packing logic itself has to branch on the cap and apply std430-style stride when supported.

This means:

• JS-side packing layout needs to be cap-aware (per-device, since Firefox/Safari may not support it)
• Shader-side struct layouts may need conditional padding for portability
• Existing uniform buffer formats / layouts may need versioning to avoid breaking pre-built shaders

Suggested approach

1. Add the WGSL feature cap infrastructure (supportsUniformBufferStandardLayout, CAPS_UNIFORM_BUFFER_STANDARD_LAYOUT, requires uniform_buffer_standard_layout;) — trivial, follows the existing pattern (#8785, #8786, #8787, #TBD).
2. Audit the JS uniform-buffer packing code paths to identify where the std140 alignment is applied and what would need to change.
3. Decide on a portability strategy: do shaders always use the tighter layout where supported and fall back where not, or is this opt-in per uniform buffer?
4. Implement and validate across the WebGPU paths.

References

• Chrome 144 release notes
• WGSL spec section on uniform buffer layout