[DRAFT] RFC Bonds orders perception

[papillot]

Description

This PR is a WIP. POC for the Mol* implementation of Roger Sayle's algorithm for bond perception https://www.daylight.com/meetings/mug01/Sayle/m4xbondage.html which is mentioned in different bug reports #36 #626 #1361

The PR is not ready yet: the code has been partly vibe coded: it requires simplifications and thorough review. Also, the validation has only started and the implementation is incomplete (inter-unit bonds are not considered as of now, behavior with ).

While working on these tasks, I would like advices about the integration in Mol* codebase:

• Parameter for bond order perception: for now it's optional, enabled by default but only kicking when there is no known bond order connectivity (not conn_bond records, or not a canonical-templated residue for which there is a table)
• tests: unit tests have been added (they will be reviewed and cleaned), but more importantly validation test. These rely on a manifest file (committed) which lists ligands codes and PDB codes to download legacy PDB and extract the ligand coordinates only as well as the CCD cif as the source of truth (not committed). As of now, all this code is buried within the src/mol-model/**/bonds source tree. Shall it move under src/tests?
• integration with the other chemistry aware code: this code assigns geometries to atom, computes aromaticity, matches some functional groups with charges, etc.. It should simplify heuristics and remove some assumptions that are made in the ValenceModel. For now, those assignments are lost after completion.

Actions

• Added description of changes to the [Unreleased] section of CHANGELOG.md
• Updated headers of modified files
• Added my name to package.json's contributors
• (Optional but encouraged) Improved documentation in docs

[arose]

this is a tricky one

validation tests should be separate from the spec tests

unless we move chemistry perception into structure creation (which would be a very big change I am not sure about) I feel this should better be calculate property like the other chemistry perception stuff and you would need to explicitly opt-in to use it instead of the given bond-orders in unit.bonds. For sure, more cumbersome but much less obtrusive.