feat(benchsdk): add optional display manifest to BenchmarkConfig and upload it on upsert

[dtice25]

Summary

Makes the *.bench.ts file the source of truth for both display and scoring metadata. This PR adds an optional display manifest to BenchmarkConfig and also makes scoring.metrics unit metadata derive from display.metrics, removing the risk that the two config blocks drift.

What changed

• packages/benchsdk-runner/src/bench-config.ts

  • Added display?: BenchmarkDisplayConfig to BenchmarkConfig.
  • BenchmarkDisplayConfig carries metrics, steps, and overview defaults.
  • defineBenchmarkConfig validates the manifest shape and rejects unknown metric display fields.
  • validateBenchmarkScoringConfig now accepts an optional display argument: scoring.metrics[i].unit is optional, and when both scoring and display declare the same key the units must match.

• packages/benchsdk-runner/src/scoring.ts

  • BenchmarkScoringMetric.unit and MetricScoring.unit are now optional.
  • scoringConfigToSpec(config, dimensions, display) resolves each scoring metric's unit from the matching display.metrics entry, falling back to metric.unit, then defaulting to an empty string.
  • lowerIsBetter / higherIsBetter accept an optional unit so existing onScore callbacks keep working.

• packages/benchsdk-runner/src/runner.ts

  • Passes config.display into scoringConfigToSpec so runtime scoring uses the canonical display unit.

• packages/benchsdk-runner/src/__tests__/bench-config.test.ts and src/__tests__/scoring.test.ts

  • Added tests for unit derivation, fallback, and conflict validation.

• benchmarks/browser/browser.bench.ts, benchmarks/sandbox/tti.bench.ts, packages/create-bench/src/index.ts

  • Removed unit from scoring.metrics and left it on display.metrics only.

Why

Previously the platform had to know about each benchmark's metrics in custom-metrics.ts and title-cased step names. Now the author declares labels, units, sort direction, and default overview layout in the same file that defines the workload, and scoring can reference those metrics without duplicating unit metadata.

Backwards compatibility

display is optional. Benchmarks without it continue to use existing platform-side fallbacks (METRIC_DISPLAY and title-cased step names). scoring.metrics[i].unit is still accepted when display is absent.

Link to Devin session: https://app.devin.ai/sessions/a45e3a08012c47e989ccfc4c055275ac
Open in Devin Desktop: https://app.devin.ai/desktop/session/a45e3a08012c47e989ccfc4c055275ac?variant=devin
Requested by: @dtice25

---

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[open-cla]

Contributor License Agreement

All contributors are covered by a CLA.