Document simplified compute shaders; add Compute Shaders Advanced page

[mvaligursky]

Stage 2 of the WGSL docs restructure (Stage 1 split the old WGSL Specifics page into WGSL Reflection, WGSL Capabilities, and WGSL Vertex and Fragment Shaders). Now that compute shaders support automatic bind-group reflection, the Compute Shaders page leads with the simplified syntax — matching how vertex/fragment shaders are documented — and manual bind-group control moves to a dedicated Compute Shaders Advanced page.

Changes:

• WGSL Reflection — intro generalized to vertex/fragment/compute; added a Storage Textures section.
• Compute Shaders — rewritten simplified-only: declare resources with no computeBindGroupFormat, loose uniforms, simplified indirect-dispatch snippet, a note that computeDispatch must run inside the render frame, and examples reordered (simplified first).
• Compute Shaders Advanced (new) — manual computeBindGroupFormat / computeUniformBufferFormats, explicit @group/@binding, the rule for mixing reflected and manual resources (reflected → group 1 when a manual format is supplied, else group 0), and a note that manual bind groups remain fully supported (migrate at your leisure).
• Sidebar — Compute Shaders Advanced added after Compute Shaders.
• Japanese — reconciled to the new structure: added wgsl-capabilities, wgsl-vertex-fragment-shaders, and compute-shaders-advanced (reusing existing translations), trimmed/generalized the JA reflection page, rewrote the JA compute page, and fixed moved-anchor links.

Notes for reviewers:

• Timing: the simplified-compute content describes behavior added by engine PR feat(webgpu): automatic bind-group reflection for compute shaders engine#8821 — merge this once that engine version ships.
• JA review: moved sections reuse existing translations, but the net-new Japanese prose (Storage Textures, the simplified compute intro, the Advanced page intro + mixing rule, the dispatch-in-frame note) is a best-effort translation and would benefit from a native-speaker pass.