feature request: make graph_objects also dataframe agnostic

I was wondering, since narwhals was such a success in the plotly express api, whether the graph_objects api could get the same treatment. For the few trace types I tried (Scatter, Bar), arguments such as x, y, marker_color etc., accept tuple, list, numpy array, or pandas Series.

It would be great to be able to pass arrow or polars series as well, via narwhals, though I have no idea how feasible that would really be.

Desired outcome would be something like this:

"""
Fresh environment
>>> uv init
>>> uv add plotly polars
"""

import plotly.graph_objects as go
import polars as pl

df = pl.DataFrame({"x": [0, 1, 2, 3, 4], "y": [0, 1, 4, 9, 16]})

fig = go.Figure()
fig.add_trace(
    go.Scatter(x=df["x"], y=df["y"]),
)
fig.show()

Currently one has to either transform to a list, pandas series, or if numpy is installed, there is some implicit transformation, which allows to pass polars series directly.

Beginne damit, das bereitgestellte polars-Beispiel über die graph_objects-API nachzustellen, mit Schwerpunkt auf go.Figure.add_trace mit Scatter und Bar. Untersuche, wie ihre Argumente x, y und marker_color derzeit Listen, NumPy-Arrays und pandas Series verarbeiten. Als erledigt gilt die Aufgabe, wenn die relevanten graph_objects-Trace-Argumente polars- und Arrow-Serien über narwhals akzeptieren, ohne eine Konvertierung zu erfordern.