Can't interchange pl.DataFrame with no columns

[honno]

Polars version checks

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of Polars.

Issue description

One cannot interchange dataframes which contain no columns (but can if they contain even one empty column). A pretty whatever issue, but thought I'd document for posterity's sake.

Reproducible example

>>> import polars as pl
>>> pl.DataFrame({}).__dataframe__()
Traceback
    ~/anaconda3/envs/df/lib/python3.10/site-packages/polars/dataframe/frame.py:1218, in DataFrame.__dataframe__(self, nan_as_null, allow_copy)
    ->  1218 return self.to_arrow().__dataframe__(nan_as_null, allow_copy)
    ~/anaconda3/envs/df/lib/python3.10/site-packages/polars/dataframe/frame.py:1839, in DataFrame.to_arrow(self)
    ->  1839 return pa.Table.from_batches(record_batches)
    ~/anaconda3/envs/df/lib/python3.10/site-packages/pyarrow/table.pxi:3972, in pyarrow.lib.Table.from_batches()
ValueError: Must pass schema, or at least one RecordBatch
>>> pl.DataFrame({"foo": pl.Series(values=[])}).__dataframe__()
<pyarrow.interchange.dataframe._PyArrowDataFrame at 0x7fc0da125c30>

Expected behavior

>>> pl.DataFrame({}).__dataframe__()
<pyarrow.interchange.dataframe._PyArrowDataFrame>  # no error

Installed versions

>>> pl.show_versions()
--------Version info---------
Polars:      0.17.14
Index type:  UInt32
Platform:    Linux-5.19.0-41-generic-x86_64-with-glibc2.35
Python:      3.10.11 | packaged by conda-forge | (main, May 10 2023, 18:58:44) [GCC 11.3.0]

----Optional dependencies----
numpy:       1.24.3
pandas:      2.1.0.dev0+779.gc96dbb77eb
pyarrow:     12.0.0  (technically using a nightly build)
connectorx:  <not installed>
deltalake:   <not installed>
fsspec:      2023.5.0
matplotlib:  3.7.1
xlsx2csv:    <not installed>
xlsxwriter:  <not installed>

[zundertj]

Thank you for reporting this, I have proposed a fix in #9144.