Add `CI_CPU_MEMORY_LIMIT_GB` to check_failed_tests workflow (#47884)

* Add CI_CPU_MEMORY_LIMIT_GB to check_failed_tests workflow

The check_failed_tests workflow re-runs failing tests on the same K8S
instance-sharing runners as model_jobs.yml, where each runner sees the
full machine CPU RAM (~750 GB) instead of its allocated share. Without
this cap, `device_map="auto"` produces a different device map than in
the original run, making the comparison inconsistent.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* Use consistent CI_CPU_MEMORY_LIMIT_GB comment

Match the comment style from transformers-ci's daily-ci_reusable_model_job.yml.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>