refactor: implement ModelBindingsMapper utilizing a DataFrame

This change introduces a refactor of the important ModelBindingsMapper
class using a pandas DataFrame for effecting grouping and aggregation
and a CurryModel utility for partially instantiating a given Pydantic
model with checks running for every partial application, allowing for
fast validation failure.

Closes #170.

Aggregation behavior is now only triggered for actually aggregated
fields, and not for top-level models as well. Therefore this also closes #181.

rdfproxy/mapper.py

@@ -16,7 +16,7 @@
 from rdfproxy.utils.utils import CurryModel, FieldsBindingsMap
 
 
-class _ModelBindingsMapper(Generic[_TModelInstance]):
+class ModelBindingsMapper(Generic[_TModelInstance]):
     """Utility class for mapping bindings to nested/grouped Pydantic models.
 
     RDFProxy utilizes Pydantic models also as a modelling grammar for grouping
@@ -125,7 +125,7 @@ def _get_unique_models(models: Iterator[_TModelInstance]) -> list[_TModelInstanc
     def _instantiate_grouped_model_from_df(
         self, df: pd.DataFrame, model: type[_TModelInstance]
     ) -> _TModelInstance:
-        """Instantiate a grouped model from a pd.DataFrame (a group dataframe).
+        """Instantiate a grouped model  pd.DataFrame (a group dataframe).
 
         This handles the GROUPED code path in _ModelBindingsMapper._instantiate_models.
         """