[BUG] cuDF-Polars string-column serialization is broken

[rjzamora]

Describe the bug
#18146 recently skipped a large number of cudf-polars tests that were being caused by this bug.

As far as I can tell, there is something going wrong when we convert a pl.DataFrame object to a cudf_polars DataFrame object when there are string columns present. For example, I am able to serialize string columns that originate from a pyarrow table, but I cannot serialize the same data if it originates from a pl.DataFrame.

Steps/Code to reproduce bug

This works...

import pylibcudf as plc
import pyarrow as pa

arrow_tbl = pa.table({"a": ["a", "bb", "ccc"]})
t = plc.interop.from_arrow(arrow_tbl)

packed = plc.contiguous_split.pack(t).release()
table = plc.contiguous_split.unpack_from_memoryviews(*packed)

This results in errors...

import polars as pl
import pylibcudf as plc
import pyarrow as pa
from cudf_polars.containers import DataFrame

pdf = pl.DataFrame({"a": ["a", "bb", "ccc"]})
t = plc.interop.from_arrow(pdf.to_arrow())

packed = plc.contiguous_split.pack(t).release()
table = plc.contiguous_split.unpack_from_memoryviews(*packed)

Additional context
This is a blocker for multi-GPU polars work.