[PM-35399] introduce benchmark for custom-elements-galore test-the-web scenario

[audreyality]

🎟️ Tracking

https://bitwarden.atlassian.net/browse/PM-35399

📔 Objective

Introduce benchmark for custom-elements-galore test-the-web scenario.

Depends upon bitwarden/clients#20600

[github-actions]

Bitwarden Claude Code Review

Overall Assessment: APPROVE

Adds a single new benchmark spec (benchmarks/static/custom-elements-galore.spec.ts) that measures shadow-DOM probe characteristics on a custom-element-heavy page from test-the-web. The file mirrors the structure of the existing rerendering-input-grid.spec.ts benchmark, passes a custom measure-name list to createBenchmarkTest (a documented API in benchmarks/README.md), and asserts instrumentation is enabled before holding the page long enough for the initial probe sweep to complete. No findings to report.

Code Review Details

No findings.

Notes considered and dropped during validation:

• The benchmark uses waitForTimeout(SETTLE_MS) rather than waiting on a selector. The README discourages fixed sleeps, but this scenario is intentionally static — there is no UI affordance to await — and the inline comment explains the wait covers the content-script collection and mutation-observer batch flush before perfCapture teardown. This matches the file's stated intent.
• The destructure only takes test (not expect). This is correct: benchmarks intentionally avoid assertions per the README ("Captures runtime metrics, doesn't assert expectations").
• The scenario URL depends on a corresponding test-the-web PR (scenarios/stability/custom-elements-galore/). Coordination of the test-site image bump is a workflow concern, not a code review finding.

[audreyality]

The test-the-web scenario will not be merged, so I'm closing this PR as well.