perf(bench): add render diagnostic scenarios

[chh-ay]

Summary

• add measured-only browser scenarios for formula-heavy paint, long unique-text scrolling, fractional device-pixel scrolling, and unresized one-million-row geometry
• thread the requested scenario list through browser execution, failure reporting, artifacts, and replay validation
• retain a clean three-round capture in bench/results/render-diagnostics.{json,md}
• keep all four scenarios out of regression gating until repeat spread supports promotion

Repeatability

Three rounds, three measured samples per round, one warmup, 100,000 rows:

• formula paint: median spread 11.64%, p95 spread 18.56%
• long-text scroll: median spread 4.16%, p95 spread 4.79%
• fractional scroll: median spread 25.18%, p95 spread 45.40%
• one-million-row geometry: median spread 3.05%, p95 spread 16.53%

Verification

• diagnostic browser smoke: 4/4 scenarios passed
• retained three-round capture: 12/12 cells passed; replay validation passed
• focused tests: 25 passed
• repository typecheck: passed
• repository lint: passed with the existing 2 MiB corpus-size warning
• full unit run: 1,279 passed / 7 failed; the same 7 failures reproduce on 0.4.0 in docs/test/compatibility-inventory.test.ts and bench/test/billion-cell-gate.test.ts

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
sheetwrite	Ready	Preview, Comment	Jul 27, 2026 2:50pm