Add offline policy evaluation module and update dependencies

 ### Changes
 * Introduced `offline_policy_evaluator.py` with classes for propensity score estimation and offline policy evaluation.
 * Updated `pyproject.toml` to include new dependencies: `bokeh`, `obp` and `optuna`. Further adjusted existing dependencies to compatible versions.
 * Added class method to PyBanditsBaseModel on base.py to allow seeing default values for arguments that were not passed to the model.
 * Added test_offline_policy_evaluator.py as a test suite for the OfflinePolicyEvaluator.
 * Added `visualize_via_bokeh` and `in_jupyter_notebook` utility functions.

pybandits/base.py

@@ -48,6 +48,10 @@ class PyBanditsBaseModel(BaseModel, extra="forbid"):
     BaseModel of the PyBandits library.
     """
 
+    @classmethod
+    def _get_value(cls, key: str, values: Dict[str, Any]) -> Any:
+        return values.get(key, cls.model_fields[key].default)
+
 
 class Model(PyBanditsBaseModel, ABC):
     """