feat(tf/pt): add/refact lammps support for spin model

[iProzd]

Summary by CodeRabbit

• New Features

  • Introduced spin handling capabilities in various models, enhancing energy and force computations.
  • Added support for new parameters in compute methods to accommodate ghost atoms and neighbor lists.
  • Implemented new classes and methods for managing spin interactions, including DeepSpin, DeepSpinModelDevi, and their TensorFlow and PyTorch counterparts.
  • Added a new pair style for LAMMPS simulations: PairDeepSpin.
  • Enhanced the DeepPot framework with spin support, including new API structures and functions.
  • Introduced a new function for writing LAMMPS data that includes spin information.
  • Added comprehensive functionality for error handling and tensor management in DeepSpinPT and DeepSpinTF.
  • Established a new testing framework for the DeepSpin class to validate its methods under various configurations.

• Bug Fixes

  • Improved error handling in TensorFlow and PyTorch integrations to ensure robustness during model execution.

• Tests

  • Established comprehensive unit tests for the DeepSpin class, validating spin-related computations and ensuring accuracy in energy and force calculations.
  • Introduced additional tests for the DeepSpin class, focusing on the compute method and verifying outputs against expected values.
  • Added unit tests for the DeepSpinPT class to validate its functionality.
  • Enhanced testing framework for LAMMPS spin simulations, ensuring correctness across various scenarios.
  • Developed tests for the DeepSpin class to validate its behavior with and without periodic boundary conditions.

• Documentation

  • Enhanced documentation for new methods and classes to clarify usage and functionality related to spin interactions.