CI workflow gaps: flaky-run traces are never uploaded and check:links never runs (needs human: workflow edits) #1188

Consolidates #1179 and #1137 — both require edits under .github/workflows/, which the Hive contributor lane cannot push, so they should land together as one human-authored CI PR.

Scope

1. Flaky-run trace upload ([quality] CI discards the traces of flaky Playwright runs: retries:1 exits 0, so the if: failure() trace upload never fires #1179): Playwright runs with retries: 1, so a spec that fails then passes on retry exits 0 and the if: failure() trace-upload step never fires. CI silently discards the evidence of every flaky run. Fix: upload traces on retry/flake as well (e.g. if: always() with retention limits, or detect flakes from the report and upload conditionally). A live example: the e2e job on PR test(mirror): cover mirrorArtworkUrls' non-regular-file destination arm #1165 failed on a change touching only a unit test file.

2. check:links never runs ([quality] check:links is configured and guarded by a test, but no workflow ever runs it #1137): check:links is configured in package.json and guarded by a unit test, but no workflow invokes it. Fix: run it in CI (scheduled and/or on PRs touching content).

Completion criteria

• Flaky e2e runs leave retrievable traces/artifacts.
• check:links runs in at least one workflow.
• One PR; closes this issue, [quality] CI discards the traces of flaky Playwright runs: retries:1 exits 0, so the if: failure() trace upload never fires #1179, and [quality] check:links is configured and guarded by a test, but no workflow ever runs it #1137.