Extend `write_parquet` and `read/scan_parquet` for `Collection` and `Schema` beyond filesystem locations

The current implementation is limited to (local) filesystem support (e.g., [here](https://github.com/Quantco/dataframely/blob/a3b90c2a1246309fcfb94cc561cf86e3bca35f93/dataframely/collection.py#L675-L678)). We should extend it to also support S3 locations or other storage backends.