#26110 ci(rescue): watch the dedicated nightly-next builder's label

ci-owned-pool-rescue.yml checks out main and runs main's scripts/ci/owned_pool_rescue.py. On feat-cmux-next, #26048 has pick-nightly-build-runner offer the dedicated nightly-next builder's label (glaeda-trusted-next-nightly) first, while that one mini is idle. Two runs can both see it idle, and the second run's build then queues on that single machine. The rescue has to recognize the label so it can move that build after the side budget, as it does on an owned pool. On rescue, the build is re-run as attempt 2 on Blacksmith. Rule: no job waits on one reserved machine.

Change:

• pr_runner_pool.NIGHTLY_DEDICATED_LABEL and nightly_dedicated() are added on main, matching feat-cmux-next's copy.
• owned_pool_rescue.job_pool() now returns that label.
• The label is not a pull request pool: persistent() is unchanged, so no PR picker offers it.

Test: tests/test_ci_owned_pool_rescue.py, 210 OK, with a new test for the label.

Order: this lands before CI_NIGHTLY_NEXT_DEDICATED_LABEL is set. Until then nothing asks for the label.

Risk: low. One more label is recognized; nothing else changes.

Agent session: agent session

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

Summary by CodeRabbit

• Bug Fixes
  • The dedicated nightly runner is now correctly recognized by CI monitoring. It remains separate from the pool used for pull requests.