Reduce flaky CI failures

[orionarcher]

Examples 5.3_Hot_Swap_WBM.py and the hugging face CI hook in the fairchem testing both appear to be a bit flaky and fail occasionally

[orionarcher]

Flakiness with graph-pes:

https://github.com/Radical-AI/torch-sim/actions/runs/14447521495/job/40511689547?pr=141

https://github.com/Radical-AI/torch-sim/actions/runs/14447521495/job/40511683512?pr=141

[orionarcher]

As does flakiness with meta-tensor:

https://github.com/Radical-AI/torch-sim/actions/runs/14460092612/job/40550984333?pr=146

[orionarcher]

And fairchem is still flaky too:

https://github.com/Radical-AI/torch-sim/actions/runs/14470883508/job/40584221809

[WillEngler]

The example logs Orion posted got garbage collected by GH Actions, so reposting some more recent logs as I try to understand the root cause:

graph-pes failues:

• https://github.com/TorchSim/torch-sim/actions/runs/18176792760/job/51744500738 (casio3_sim_state)
• https://github.com/TorchSim/torch-sim/actions/runs/18144062426/job/51641704586#step:8:149 (rattled_sio2_sim_state)
• https://github.com/TorchSim/torch-sim/actions/runs/17803499812/job/50609353292 (tio2_sim_state)

can't download sevennet weights from GitHub? (less common than the graph-pes stuff):

• https://github.com/TorchSim/torch-sim/actions/runs/16889250592/job/47844588700