Support polars directly in DataGrid and render.data_frame

[machow]

Currently, the DataGrid supports polars by calling its polars.DataFrame.to_pandas() method. From pairing with @schloerke, I can take a pass at supporting polars directly in the DataGrid. This will give me a chance to see what DataFrame specific logic is in there.

TODO:

• Separate out DataFrame specific logic (similar to in Great Tables _tbl_data.py)
• Should not create hard dependencies on any given DataFrame library
• Should handle polars dtypes (a big part of the DataFrame specific logic right now).
• After submitting PR, all the DataFrame-specific logic should be in one place, so I can evaluate plugging in narwhals

Example app

from shiny.express import render
import polars as pl

df = pl.DataFrame({"x": [1,2,3], "y": ["a", "b", "c"], "z": [{"val": 1}, {"val": 2}, {"val": 3}]})

@render.data_frame
def frame():
    return render.DataGrid(df)

NOTES:

• htmltools.HTML subclasses str, but this causes polars.Series to coerce it to str. This means that you can't get HTML back. (subclassing UserString or something avoids this behavior).

[schloerke]

Fixed in #1474