feat(anari): add interactive ray tracing budget

[ibgreen-openai]

Goals

• Make ANARI ray tracing interactive by default on CORE WebGPU.
• Preserve animated scenes, HDR presentation, and application-owned GPU command submission.

Changes

• Add half-resolution defaults, adaptive 30 FPS budgeting, quantized scale changes, and 1/2/4 interleaved pixel coverage.
• Split GPU acceleration and trace graphs so camera-only and lighting-only frames reuse retained BVH buffers.
• Add cached triangle extraction, stable instance identities, previous transforms, temporal reprojection, depth/normal rejection, neighborhood clamping, and per-pixel sample counts.
• Add bounded rotating shadow-light sampling with shadowSamplesPerFrame: 0 as the all-lights opt-out.
• Expose ray-tracing quality controls and diagnostics through ANARI types, strict schemas, both examples, and docs.

Verification

• node_modules/.bin/tsc --build modules/*/tsconfig.json --pretty false
• Focused ANARI + experimental node suite: 19 files / 116 tests passed.
• ANARI showcase production build: 1,653 modules transformed.
• Targeted Biome checks and git diff --check.
• Real CORE WebGPU browser run: 312 instances, one draw, adaptive 75% → 50% internal resolution, settled around 23 ms/frame, no app-side WebGPU validation errors.

Limits

• Object/instance BVH traversal is source-ordered; triangles inside surviving meshes remain linear.
• No hardware ray tracing, Morton ordering, per-mesh BVHs, indirect bounces, denoising, or volume tracing yet.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 25582a9e2e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".