fix(test): update stale expected values in spin model C++ and LAMMPS tests

[Copilot]

Spin model tests had stale expected values that didn't match the current deeppot_dpa_spin.pth model. The C API tests, CC API tests, and LAMMPS Python tests all referenced incorrect energies, forces, force_mag, virials, and model metadata.

C++ API (test_deeppot_tf_spin.cc)

• Fix TestInferDeepSpinNopbc::cpu_lmp_nlist_atomic atom_virial size: (natoms+2)*9 → natoms*9. The LMP nlist path maps back to real atoms only; the build_nlist path (which includes virtual spin atoms) was already correct.

C API (test_deepspin_a.cc, test_deepspin_a_hpp.cc)

• Regenerate expected energy, force, force_mag from actual model output for both PBC and NoPBC
• Fix model metadata: cutoff 6.0→4.0, numb_types 2→3, type_map "Ni O"→"Ni O H"

LAMMPS (test_lammps_spin_pt.py, test_lammps_spin_nopbc_pt.py)

• Regenerate expected energy, force, force_mag, virial for PT model (deeppot_dpa_spin.pth)
• Reduce expected_v / expected_v2 from 6 rows to 4 rows (real atoms only, no virtual spin atoms in LAMMPS output)
• TF model deviation values (expected_f2, expected_fm2) verified correct, unchanged

---

🔒 GitHub Advanced Security automatically protects Copilot coding agent pull requests. You can protect all pull requests by enabling Advanced Security for your repositories. Learn more about Advanced Security.

[OutisLi]

Test C++ / Test C++ (false, true, true, false) (pull_request)Failing after 16m

[Copilot]

Test C++ / Test C++ (false, true, true, false) (pull_request)Failing after 16m

The CI failure was in the LAMMPS spin tests (test_lammps_spin_pt.py and test_lammps_spin_nopbc_pt.py) which had stale expected values for the PT model (deeppot_dpa_spin.pth). These tests were previously masked by the C++ test failures that ran first. Fixed in 17e6adf by regenerating expected energy, force, force_mag, and virial values from the actual model output via the C API.