`icicle`, `treemap`, and `sunburst` won't work with cuDF when using `path` parameter (master branch)

[LiamConnors]

I was running through some testing for cuDF and noted these traces will not work because they use iter_rows() when path parameter is used on the px function.

import plotly.express as px
import numpy as np
df = px.data.gapminder().query("year == 2007")
df = cudf.DataFrame(df)
fig = px.sunburst(df, path=['continent', 'country'], values='pop',
                  color='lifeExp', hover_data=['iso_alpha'],
                  color_continuous_scale='RdBu',
                  color_continuous_midpoint=np.average(df['lifeExp'], weights=df['pop']))
fig.show()

[/usr/local/lib/python3.10/dist-packages/plotly/express/_core.py](https://localhost:8080/#) in _check_dataframe_all_leaves(df)
   1859     ).get_column("null_mask")
   1860 
-> 1861     for row_idx, row in zip(
   1862         null_indices_mask, null_mask.filter(null_indices_mask).iter_rows()
   1863     ):

[/usr/local/lib/python3.10/dist-packages/narwhals/series.py](https://localhost:8080/#) in __iter__(self)
   2691 
   2692     def __iter__(self: Self) -> Iterator[Any]:
-> 2693         yield from self._compliant_series.__iter__()
   2694 
   2695     @property

[/usr/local/lib/python3.10/dist-packages/narwhals/_pandas_like/series.py](https://localhost:8080/#) in __iter__(self)
    761 
    762     def __iter__(self: Self) -> Iterator[Any]:
--> 763         yield from self._native_series.__iter__()
    764 
    765     @property

[/usr/local/lib/python3.10/dist-packages/cudf/utils/utils.py](https://localhost:8080/#) in __iter__(self)
    240         information.
    241         """
--> 242         raise TypeError(
    243             f"{self.__class__.__name__} object is not iterable. "
    244             f"Consider using `.to_arrow()`, `.to_pandas()` or `.values_host` "

TypeError: Series object is not iterable. Consider using `.to_arrow()`, `.to_pandas()` or `.values_host` if you wish to iterate over the values.

[gvwilson]

Let's document this in the section where we tell people which dataframes we support ("For example, blah blah doesn't work with cuDF yet…" - nothing fancy).

[MarcoGorelli]

thanks for spotting this, I think it should be possible to rewrite this without using iter_rows