Separate logic of ontology_id and other registries better

[Zethson]

Our current implementation relies on attribute checks (hasattr for ontology_id, _ontology_id_field, _name_field) to differentiate ontology registries from other BioRecord registries. This approach lacks robustness as unique identifier fields vary across registry types (some ontologies use ontology_id while Genes use ensembl_gene_id or stable_id depending on organism). We need a more explicit architecture to properly distinguish between registry types and their unique constraints.