Benchmark fixture trees need scoped Biome policy

Problem

Root pnpm lint:fix was applying the full Plate app lint profile to benchmarks/editor/**, which is a mix of benchmark harness code, cross-framework templates, generated-style fixture code, and static demo assets. That made an unrelated docs/rules task look red.

Symptoms

• pnpm lint:fix failed after an editor-test-harvester rules update.
• Biome reported 202 errors, mostly under benchmarks/editor/**.
• Diagnostics included CSS specificity ordering, framework-template style preferences, unused fixture variables, console usage in test runners, and a11y rules on non-product benchmark markup.

What Didn't Work

• Treating the failures as evidence that the editor-test-harvester change was wrong. The changed rule/docs files were not the owner.
• Manually editing every framework fixture. That would churn templates and static assets just to satisfy app-oriented rules that do not fit the fixture surface.
• Ignoring all of benchmarks/editor/**. That would hide real correctness issues in benchmark scripts.

Solution

Add a scoped biome.jsonc override for benchmarks/editor/** that disables the rules that are noise for benchmark fixtures and templates, while leaving real JavaScript correctness rules active.

Then fix the remaining concrete script errors directly. In this case, after the override only two Number.parseInt radix errors remained, so benchmarks/editor/scripts/utils.js now passes an explicit radix:

errors: Number.parseInt(match[2], 10) || 0,
warnings: Number.parseInt(match[3], 10) || 0,

Verification:

pnpm lint:fix

Why This Works

The benchmark tree is not the same kind of source as packages, docs app code, or registry examples. It intentionally contains many framework styles and fixture patterns. A scoped override keeps the root closeout gate useful without forcing every benchmark fixture to mimic the main app style guide.

Keeping correctness rules active unless they are fixture-specific also avoids turning the override into a blanket lint bypass.

Prevention

• When root lint fails under a benchmark/template tree, first classify the owner: product code, generated fixture, framework template, static asset, or script.
• Prefer a scoped Biome override for fixture/template policy mismatches.
• Still fix concrete correctness bugs in benchmark scripts instead of suppressing them.
• Do not use root lint debt under benchmark fixtures to block unrelated docs/rules changes once the scoped policy is green.

Related Issues

• Slate v2 tooling port should stage package build owners separately from repo-wide source lint