Add arrow schema extraction dispatch

[galipremsagar]

After #9131 was merged, it brought in a change to default parameter schema from None to "infer", this results in a failure when we try to use parquet writer in dask_cudf because of pandas specific code. This PR adds a pyarrow_schema_dispatch that will handle the pyarrow schema extraction for their respective backends.

• Closes #xxxx
• Tests added / passed
• Passes pre-commit run --all-files

[charlesbluca]

Can you give a minimal reproducer of the dask-cudf failure itself? Might help to add a gpuCI test in this PR to ensure that a breaking change like this isn't made in the future

[galipremsagar]

Can you give a minimal reproducer of the dask-cudf failure itself? Might help to add a gpuCI test in this PR to ensure that a breaking change like this isn't made in the future

Here is the minimal repro:

df = cudf.DataFrame(
            {
                "a": ["abc", "def"],
                "b": ["a", "z"],
            }
        )
ddf = dask_cudf.from_cudf(df, 3)
ddf.to_parquet("sample.parquet")

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
../envs/cudfdev/lib/python3.9/contextlib.py:79: in inner
    return func(*args, **kwds)
python/dask_cudf/dask_cudf/core.py:278: in to_parquet
    return to_parquet(self, path, *args, **kwargs)
../envs/cudfdev/lib/python3.9/site-packages/dask-2022.3.0+153.g256622943.dirty-py3.9.egg/dask/dataframe/io/parquet/core.py:844: in to_parquet
    i_offset, fmd, metadata_file_exists, extra_write_kwargs = engine.initialize_write(
../envs/cudfdev/lib/python3.9/site-packages/dask-2022.3.0+153.g256622943.dirty-py3.9.egg/dask/dataframe/io/parquet/arrow.py:516: in initialize_write
    inferred_schema = pa.Schema.from_pandas(
pyarrow/types.pxi:1492: in pyarrow.lib.Schema.from_pandas
    ???
../envs/cudfdev/lib/python3.9/site-packages/pyarrow/pandas_compat.py:525: in dataframe_to_types
    values = c.values
../envs/cudfdev/lib/python3.9/contextlib.py:79: in inner
    return func(*args, **kwds)
../envs/cudfdev/lib/python3.9/site-packages/cudf/core/single_column_frame.py:118: in values
    return self._column.values
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = <cudf.core.column.string.StringColumn object at 0x7f6373e9f1c0>
[
  "cat",
  "dog"
]
dtype: object

    @property
    def values(self) -> cupy.ndarray:
        """
        Return a CuPy representation of the StringColumn.
        """
>       raise TypeError("String Arrays is not yet implemented in cudf")
E       TypeError: String Arrays is not yet implemented in cudf

../envs/cudfdev/lib/python3.9/site-packages/cudf/core/column/string.py:5321: TypeError

I've added a GPU test for the same in this PR.