Use `ww.init_series(schema=schema)` to retain origin in partial dependence

- Currently, we explicitly pass along the origin and logical type when initializing woodwork on the updated partial dependence column. This means that if predictions ever become dependent on another woodwork attribute, we have to remember to add it at partial dependence, or we'll get incorrect predictions. To avoid that, we can pass along the original column's woodwork schema, but we have to wait for that ability to be implemented in `ww.init_schema` in [this woodwork issue](https://github.com/alteryx/woodwork/issues/1573).

Once that issue is completed and released, we should be able to do the following in partial dependence:

```python
X_eval.ww[variable] = ww.init_series(
part_dep_column,
schema=X_eval.ww.columns[variable]
)
```