DockM8 Scoring function for Thomson Sampling routine

[HiteSit]

Hi developers,

DockM8 (or a low-computational-power version of it) could be integrated as a scoring function for a [Thomson Sampling approach by PatWalters](https://github.com/PatWalters/TS) or [Thomson Sampling Enhanced](https://github.com/WIMNZhao/TS_Enhanced). This integration could benefit all Cheminformatics and Medicinal Chemistry routines that require screening an ultra-large chemical space derived from a single SMARTS pattern, without needing to enumerate the reaction. This approach would complement machine learning-enhanced docking.

While writing the scoring function might not be very complex (as it's possible to create a class that returns a float—the score—to reweight the Bayesian coefficients of Thomson Sampling), the main challenge would be making it computationally efficient. Only high-throughput docking engines like Plants or QVINAW might be suitable for this purpose.

The advantage is clear: deliver DockM8 not only to computational chemistry specialists but also to medicinal chemists (something you're already doing from day zero by writing the Streamlit interface).