[abhijeetgangan]

Summary

Closes #592, closes #615.

• Triclinic minimum_image_displacement (minimum_image_displacement only valid for orthogonal cells #592). Rounding fractional coordinates only finds the minimum image in orthogonal cells.
  • Now also checks the neighboring periodic images and keeps the shortest. Ties keep the rounded image, so orthogonal results are unchanged.
  • Matches a brute-force search on ~3,000 random Niggli-reduced cells. main returns the wrong image for 13–27% of displacements in primitive FCC/BCC/HCP cells.
  • The issue's example is a new case in test_minimum_image_displacement.
• Ewald/PME real-space truncation (EwaldModel and PMEModel truncate the real-space sum when cutoff is below the estimated Ewald cutoff #615). Option 2 from the issue: alpha is derived from cutoff and accuracy with nvalchemiops' public estimators.
  • Rattled 512-atom NaCl, default cutoff: energy error 3.2 eV → 1.4e-4 eV.
  • test_slab_matches_2d_lattice_sum had the same truncation, hidden by a loose tolerance. It now uses atol=1e-4, rtol=0.

Checklist

Before a pull request can be merged, the following items must be checked:

• Doc strings have been added in the Google docstring format.
• Run ruff on your code.
• Tests have been added for any new functionality or bug fixes.

[lil-lon]

Thanks for fixing #615! Overall, LGTM. I left minor suggestions and comments.