Consider using `ray.data.read_parquet_bulk` when possible

[jaidisido]

Is your idea related to a problem? Please describe.
The ray.data.* module has a read_parquet_bulk which is optimised to read a large number of small files. It expects a list of paths instead of a directory.
https://docs.ray.io/en/latest/data/api/input_output.html#parquet

The issue with it however is that it does not use the ParquetDataset abstraction to infer metadata.

P.S. Please do not attach files as it's considered a security risk. Add code snippets directly in the message body as much as possible.