Add measurable 10M trace contract

[ibgreen-openai]

What changed

• define stable 250K, 1M, 4M, and 10M trace workload contracts
• report exact source-buffer/device-limit compatibility and persistent GPU allocation metrics
• feed trace workload, visibility, interaction-mode, and graph-encoding counters into the reusable GPU command graph inspector
• add node and real-WebGPU regression coverage for the contract and telemetry

Why

The trace viewer can demonstrate very large workloads, but lacked a stable, measurable contract for comparing interaction cost and identifying the actual 10M-span constraint. This makes the current monolithic allocation limit explicit and provides the observability needed to evaluate chunk-native storage in the next tranche.

Impact

The experimental trace example now shows persistent allocation size, largest buffer, device-fit status, and live command-graph counters without changing rendering semantics.

Verification

• direct full module build passed
• focused TypeScript check passed
• GPU trace node tests passed (13/13)
• Chromium/WebGPU trace test passed (1/1)
• targeted Biome formatting and git diff --check passed

The repository yarn lint fix, yarn build, and yarn test wrappers could not run locally because the dependency registry returned HTTP 403 while installing the newly referenced dev-tools packages, leaving the ocular-* commands unavailable. CI is expected to exercise the complete repository gates.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 9928e7b5b8

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".