ci: auto-rerun daily pipeline on runner allocation failure

[costajohnt]

Summary

• Adds .github/workflows/auto-rerun-allocation-failure.yml which listens for completed runs of the Daily Trading Pipeline and reruns failed jobs once if the failure annotation matches GitHub's runner-pool allocation message ("not acquired by Runner of type hosted...").
• Other failure modes (test failures, script errors, real bugs) are left alone so they continue to surface normally.
• Guarded to run_attempt == 1 to prevent rerun loops if the rerun itself fails.

Context

Run #25385462949 failed at 15m4s with the annotation "The job was not acquired by Runner of type hosted even after multiple attempts", a transient GitHub infrastructure incident where the hosted runner pool never picked up the job. The pipeline self-recovered on the next scheduled tick, but the missed slot was lost. This workflow handles that class of failure automatically.

Verified the detection logic against run #25385462949: workflow_run conclusion is failure, run_attempt is 1, and the matching annotation is present on the job.

Test plan

• Confirm workflow appears under Actions tab after merge
• On next runner allocation failure (rare), verify a rerun is triggered automatically
• Verify normal pipeline failures (e.g. a test failure) do NOT trigger a rerun