Consider implementation of torque clustering

[sreichl]

• read paper
  • https://app.paperpile.com/my-library/Yang-and-Lin-2025-p6t7FzOO8B5KyQnISfpU_Fw
  • https://www.computer.org/csdl/journal/tp/5555/01/10856563/23Saifm0vLy
  • https://ieeexplore.ieee.org/document/10856563
• check code (MATLAB?! wtf) https://github.com/JieYangBruce/TorqueClustering
• decide/test if universally better than UMAP + Leiden
• ?implement by porting code to Python or Rust using AI into a separate dedicated repo that can be installed via conda and to be used within the snakemake workflow