ci: remove unstick-prs.yml; armed BEHIND PRs are already advanced by the maintainer's sweep

[askalf]

Removes .github/workflows/unstick-prs.yml.

• It reads a PR_SWEEP_TOKEN secret that was never set on this repo, so all 94 runs in the last 14 days (cron 17,47 * * * *) exited at its own guard, PR_SWEEP_TOKEN is not set — nothing to do, and reported green.
• The job it describes, running gh pr update-branch on a PR that is armed for auto-merge and BEHIND, is already done for this repo by the maintainer's own sweep, which covers every public askalf repository with a user identity. That sweep exists partly because of fix(hook): block a pinned skill the gate could not verify, in both modes #111–fix(lock): serialize read-modify-write so concurrent pins stop losing entries #115 here.

Nothing else in the repo references the workflow or the secret.

[sprayberry-secondread]

Automated review from the Sprayberry Labs fleet code reviewer.

Reviewed by the Claude second-opinion lane (second opinion, non-gating; the gating review is posted separately).

No blocking issues found in this workflow-only deletion.

Read the full 149-line deletion of .github/workflows/unstick-prs.yml. The removed workflow has schedule and manual-dispatch entry points, not a pull-request check, and its missing-token guard exits before listing or updating any PR. The latest run I inspected (35891960798) actually emitted PR_SWEEP_TOKEN is not set — nothing to do, supporting the claim that this currently does no branch-update work.

What's good: the change is narrowly scoped, removes an idle scheduled job without changing application code or CI test definitions, and explains where branch-update responsibility now lives. All reported PR checks pass, including Node 20/22 across Linux, macOS, and Windows. No local test suite was run.

Scope limit: I did not independently verify the external maintainer sweep, the complete 94-run history, or a repository-wide reference search. Removing this file also removes its manual-dispatch fallback; continued automatic branch advancement depends on the external sweep described in the PR.

[sprayberry-redline]

Automated review from the Sprayberry Labs fleet code reviewer.

Reviewed by the GPT gating lane (gating review).

APPROVE: no blocking findings in the workflow removal.

Reviewed the complete 149-line deletion, PR description, and commit messages at 45182cd4985c5d5f487f902a3ed2e0f5d6478884. The deleted workflow has only scheduled and manual triggers; it is not a reusable workflow or a pull-request test job. Removing it does not modify application code or the remaining test workflows. Its missing-token guard confirms the described successful no-op behavior when the secret is absent.

All reported CI checks pass, including the Node 20/22 test matrix on Linux, macOS, and Windows, CodeQL, triage, and pinned-skills verification. This CI-only change is exempt from the separate verification requirement; no local suite was run. I did not independently audit the external maintainer sweep or the historical secret configuration.

What's good: the change is limited to removing one redundant automation entry point, without unrelated edits.