[FEA] Properly interface with `dask` `object` -> `arrow[string]` conversions

**Is your feature request related to a problem? Please describe.**
`dask` has a behavioral change when `pandas-2.x` & `pyarrow` are installed. Some APIs try switching `object` type columns to `arrow[string]` for efficiencies arrow strings provide over object types in pandas. This causes a lot of pytest failures in `dask_cudf`, that is currently being prevented by configured a `dask.config`: `dataframe.convert-string` to `False`. However, this only happens properly for `dask.DataFrame`'s and not for `dask_cudf.DataFrame`'s.

**Describe the solution you'd like**
We need to update `cudf` & `dask` upstream to get this support for `dask_cudf.DataFrame` as well when the device object is converted to host objects.