[CI] Fix the CI jobs that fail on the OSDC images

[huydhn]

• Install python3-dev in the procgen, safety-gymnasium, minari and sglang setups, because the OSDC image's Python has no headers and building torchrl fails on Python.h.

• Stop the habitat setup from putting /usr/bin first in PATH for the rest of the script, because later pip3 calls then hit the system pip, which refuses to install on the Ubuntu 24.04 image. Also install libglib2.0-0, which OpenCV needs and that image lacks.

• Install cmake in the vllm and sglang setups, because the pinned tensordict builds a C++ extension and the OSDC image has no cmake. vllm takes it from apt, because pip builds tensordict in an isolated environment where the cmake wheel's launcher cannot import its own package.

• Move [Refactor] Bound GPU test parallelism and reap strays #4532's stray reaping into helpers/reap_strays.sh and run it on exit in the craftground and libero tests too, because their background Xvfb keeps the job's output open and the job hangs until it times out.

• Treat the killed child as gone once it is a zombie in test_run_example_terminates_timeouts (folded in from [CI] Count a zombie as reaped in test_run_example_terminates_timeouts #4547). On OSDC the job container's PID 1 is tail -f /dev/null, which never reaps orphans, so os.kill(pid, 0) keeps succeeding on the zombie.

• Map the container's hostname to loopback in the olddeps tests, as linux/scripts/run_all.sh already does, because the gloo and nccl transport tests in TestRayRB and test_inference_server.py fail with gaierror resolving it.

• Install python3-venv in the vllm, sglang, openx and minari setups, the jobs that download from the HF Hub, because linux_job_v3's HF cache sync falls back to installing the AWS CLI in a venv and the Ubuntu 24.04 image can't create one, so ci-refresh-hf-cache runs never upload the models they download.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/4544

• 📄 Preview Python docs built from this PR

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

✅ You can merge normally! (2 Unrelated Failures)

As of commit 456836b with merge base 48a06fc ():

BROKEN TRUNK - The following jobs failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• Unit-tests on Linux / tests-olddeps (3.10, 11.8, remainder) / linux-job (gh) (trunk failure)
  test/test_trainer.py::TestPPOFromEnv::test_training_and_checkpoint_resume[True-True]
• Unit-tests on Linux / tests-olddeps (3.10, 11.8, transforms) / linux-job (gh) (trunk failure)
  test/transforms/test_action_transforms.py::TestClosedLoopMultiAction::test_feedback_and_state[group_key1-3]

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

[huydhn]

@pytorchbot drci

[huydhn]

@pytorchbot drci

[huydhn]

@pytorchbot drci