Vina preparation fabricates zero charges and name-based atom types

[dprada]

What — Current preparation writes zero partial charges, guesses AutoDock types from names, and drops hydrogens by name, including polar receptor hydrogens.
How — Replace these assumptions with validated MolSysMT parameters and a documented PDBQT projection; until available, reject incomplete inputs with actionable diagnostics.
Why — The generated PDBQT can be syntactically accepted while misrepresenting the prepared receptor or ligand chemistry.
Record — devguide/pending_bugs/vina_preparation_fabricates_zero_charges_and_name_based_atom_types.md
Related — #3. DockingMT needs MolSysMT charge and atom-typing capabilities tracked in uibcdf/molsysmt#221 and uibcdf/molsysmt#222; the local Vina opt-in is provisional while those provider issues remain open.

[dprada]

Progress in 61e5f7b (Core MVP Gate C1):

• Vina now rejects DockingMT-generated zero-placeholder charges and heuristic AutoDock types by default, whether the input is a MolSysMT system or a prebuilt PreparedLigand/PreparedReceptor.
• Exploratory runs require explicit VinaProtocol(allow_provisional_preparation=True); protocol serialization and result provenance retain that choice and the preparation assessment.
• External PDBQT inputs retain an unassessed chemistry assessment.
• Guard: tests/test_engines.py::test_vina_rejects_provisional_chemistry_from_automatic_and_prepared_inputs.
• Local gates: Ruff lint/format, 61 pytest tests, devguide index check passed.

The local report is now partial: devguide/pending_bugs/vina_preparation_fabricates_zero_charges_and_name_based_atom_types.md. This issue remains open. Validated protein and ligand parameterization, scoring compatibility, and scientific preparation tests are still required; see uibcdf/molsysmt#221 and uibcdf/molsysmt#222.