Add support for path-like file name in to_csv and to_parquet

[leostimpfle]

This PR adds support for os.PathLike output filenames in the DuckDBPyRelation.to_csv and DuckDBPyRelation.to_parquet methods.

Currently, only plain Python strings are supported which requires explicit casting of pathlib.Path objects.

from pathlib import Path
import duckdb

# On main this throws TypeError: to_csv(): incompatible function arguments.
duckdb.sql("SELECT 43").to_csv(Path("test.csv"))

# Support for plain strings is unaffected
duckdb.sql("SELECT 43").to_csv("test.csv")

# Other arguments throw: _duckdb.InvalidInputException: Invalid Input Error: Expected a str, bytes, or os.PathLike object for the file path, not 'int'
duckdb.sql("SELECT 43").to_csv(123)

The core change is the addition of PathToString which is used in DuckDBPyRelation::ToCSV and DuckDBPyRelation::ToParquet. I have added tests in tests/fast/api/test_to_csv.py and tests/fast/api/test_to_parquet.py for the new argument types.

@evertlammerts Please let me know any comments. I have a few further ideas on similar small-ish improvements to duckdb-python which I would love to work on. (Please let me know if I should always first open discussions before PRs.)

[NickCrews]

nit: rename to TryEnsurePathString ("Ensure" implies that it can handle an already-string, "Decode" makes it sound like it can only handle Paths)

[NickCrews]

LGTM, two small nits that are non blockers for me (I'm not a maintainer though)

[NickCrews]

nit: rename to EnsurePathString (or whatever you end up with for the function above, without the "Try" prefix)