Modular functions, support grouping chains, and partial support for Chai-1 models

[y1zhou]

New

• Packaged the script into an installable file structure.
• Added .pre-commit-config.yaml to standardize formatting and catch linter errors.
• Added support for specifying the model type from CLI arguments and skipping the guessing process.

Changed

• Refactored global variables and subroutines in the original script into modular functions. Notably, there is a ipsae() function that can be imported by other scripts for downstream processing. Results are stored as lists of dataclasses, which can be trivially transformed to pandas or polars data frames.
• Vectorized inner loop calculations where possible for performance improvements.
• Improved CLI argument parsing, which now supports outputting results to stdout and to a specific directory via --output-dir.

Fixed

• Chain index parsing logic from Fix chain index for iptm matrix #19
• Redundant iterations of the input arrays for calculating different scores.
• Counterintuitive row ordering in the generated PyMOL script.

These changes collectively improve the usability, reproducibility, and maintainability of the IPSAE scoring tool.

[y1zhou]

Since commit 1dd0697, two additional things were added:

• CLI options --chain-groups to specify groups of chains that would be merged during score calculation. For example in an antibody-antigen complex scenario, H+L/A would compute interface scores for the two antibody chains vs. the antigen chain.
• Unit tests for the chain group parsing logic, and regression tests using tables generated from the original script to ensure my changes did not affect the output scores of individual chains.