chore(chromium): switch linux arm64 to cft builds

[Dmitry Gozman (dgozman)]

Bots: https://github.com/microsoft/playwright-browsers/actions/runs/33159155304
Fixes#42386.

[github-actions]

Test results for "tests 1"

2 flaky

⚠️ [chromium-library] › library/video.spec.ts:699 › screencast › should capture full viewport on hidpi `@chromium-ubuntu-22.04-node20`
⚠️ [firefox-page] › page/page-emulate-media.spec.ts:144 › should keep reduced motion and color emulation after reload `@firefox-ubuntu-22.04-node20`

51251 passed, 1240 skipped

---

Merge workflow run.

[github-actions]

Test results for "MCP"

4 failed
❌ [chromium] › mcp/clipboard.spec.ts:19 › clipboard write without permission dialog @mcp-macos-latest-chromium
❌ [chromium] › mcp/clipboard.spec.ts:19 › clipboard write without permission dialog @mcp-windows-latest-chromium
❌ [chromium] › mcp/clipboard.spec.ts:19 › clipboard write without permission dialog @mcp-ubuntu-latest-chromium
❌ [firefox] › mcp/annotate.spec.ts:386 › should cancel browser_annotate when the MCP request is aborted @mcp-windows-latest-firefox

8255 passed, 1361 skipped

---

Merge workflow run.

[github-actions]

Hi, I'm the Playwright bot and I took a first look at the CI failures.

🟢 CI is clear — all failures are pre-existing flakes

This PR only rewrites the Chromium linux-arm64 download and executable paths to CFT builds. None of the failing tests run on arm64, so the diff can't reach them. Both failures also flip verdict across hundreds of unrelated runs in the CI results DB — they're known flakes.

Details

Overall: the diff is scoped to registry/index.ts arm64 entries for chromium/chromium-headless-shell. The failing MCP tests ran on macOS / Windows / ubuntu-latest (x64) Chromium and Windows Firefox — never on arm64 — so nothing here is caused by this PR.

Pre-existing flake / infra

• [chromium] › mcp/clipboard.spec.ts:19 › clipboard write without permission dialog (@mcp-macos-latest-chromium, @mcp-windows-latest-chromium, @mcp-ubuntu-latest-chromium) — flaky. Across the results DB the chromium project failed 8 of 730 runs (~1%) and passed 722, on SHAs unrelated to this PR; chrome, msedge and webkit pass 100%. The clipboard permission flow has nothing to do with arm64 binary fetching.

• [firefox] › mcp/annotate.spec.ts:386 › should cancel browser_annotate when the MCP request is aborted (@mcp-windows-latest-firefox) — flaky. It flips verdict across projects: firefox failed 6/721, chrome 6/734, msedge 4/720, all passing the vast majority of runs on unrelated SHAs. A request-abort race, and a Firefox failure can't come from a Chromium-only change anyway.

The two entries in the "tests 1" report (video.spec.ts:699, page-emulate-media.spec.ts:144) were flaky (passed on retry), so there's nothing to triage there.

_{Triaged by the Playwright bot - agent run}