feat(examples): zarr dataset configs and .pdmsh converter for the unified external aero recipe

[negin513]

Description

Split out of #1798 per review discussion (scope proposal): the recipe integration for the mesh-zarr storage path, held as a separate decision from the library core until the team settles whether to migrate curated datasets.

What's included

• drivaer_ml_surface_zarr.yaml / drivaer_ml_volume_zarr.yaml: DrivAerML dataset configs reading physicsnemo-domainmesh-zarr groups. They differ from their .pdmsh twins only in the reader: block — the full transform stack and all five recipe models run unchanged.
• src/convert_pdmsh_to_zarr.py: transitional CLI for datasets already curated to .pdmsh (new curation should emit the schema directly from PhysicsNeMo-Curator; companion PR: feat(mesh): DomainMeshZarrSink emitting the physicsnemo mesh-zarr schema physicsnemo-curator#64).
• dataset_paths.yaml: adds the drivaer_ml_zarr path key.

Evidence (carried over from #1798, measured with these configs)

• Matched training A/Bs (seed 42): surface — final train loss 0.1266 (.pdmsh) vs 0.1265 (zarr); volume — 0.2648 vs 0.2680 train / 0.0998 vs 0.0987 val at epoch 3; trajectories track epoch-for-epoch within subsample-RNG noise.
• End-to-end: GeoTransolver surface and volume both train from zarr through the recipe's train.py (volume incl. on-the-fly SDF against the in-group full-resolution STL boundary).
• DataLoader (benchmark_io, 8 DrivAerML cases, Lustre): 3.66 cold / 7.96 warm batches/s (.pdmsh) vs 9.23 / 23.64 (zarr) — raw logs.

Dependencies

Merge-blocks on #1798 (imports save_domain_mesh_to_zarr / to_cell_soup; configs target ${dp:ZarrMeshReader} / ${dp:ZarrDomainMeshReader}). Draft until #1798 lands and the dataset-migration decision is made.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes (config-only + CLI; exercised end-to-end by the A/Bs above and the library tests in Feat(datapipes): Zarr mesh schema and readers — a zarr storage path for mesh-native CAE pipelines #1798).
• The documentation is up to date with these changes.
• The CHANGELOG.md is up to date with these changes.
• An issue is linked to this pull request.

🤖 Generated with Claude Code

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.