[CI] Move benchmarks to the nightly run, label opt-in on PRs

[vmoens]

Stack from ghstack (oldest at bottom):

• [Doc] Cache executed sphinx-gallery output between doc builds #3983
• -> [CI] Move benchmarks to the nightly run, label opt-in on PRs #3982
• [CI] Demote the Windows suite to nightly #3981
• [CI] Shard the SOTA smoke tests into two parallel jobs #3980
• [CI] Contain hard hangs in the optdeps job #3979
• [CI] Shard the olddeps job #3978
• [CI] Run the CPU quarantine shards on the oldest and newest Python only #3977
• [CI] Fan out the tests-cpu serial quarantine into parallel jobs #3976

Benchmarks cost 2 x ~40-60 min GPU jobs on every push to main and on
every PR push, while their only consumers are trend dashboards.

• benchmarks.yml: drop the push-to-main trigger. The nightly
  orchestrator's daily workflow_call becomes the sole feed of the
  gh-pages trend: its skip-upload flag now derives from the run mode
  (upload in "full" mode only, so dry-run and smoke-test runs cannot
  touch gh-pages). workflow_dispatch remains for on-demand runs.
• benchmarks_pr.yml: opt-in via the "benchmarks" PR label, with the
  "labeled" trigger type so applying the label starts the comparison
  immediately.
• benchmarks_pr_comment.yml: exit cleanly when the benchmark job was
  skipped and no artifacts exist, instead of failing on an empty
  download.

Accepted tradeoffs: the trend granularity drops from per-push to
daily, and the 200% regression alert now implicates a day of commits
rather than a single push.

Co-Authored-By: Claude Fable 5 noreply@anthropic.com

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/3982

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

⏳ No Failures, 24 Pending

As of commit f1514b9 with merge base a74b478 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.