Create load from parquet

[PhilippeMoussalli]

PR that creates a generic component to load data stored as a parquet from remote storage

Largely based on the load from hub component with some modification. Currently needed for the datacomp pipeline.

Future modification could include loading data from local storage though we'd need to make sure that the data is mounted locally

[mrchtr]

The component does probably the job for now.
I am just curious about if we can reuse the load from hub component completely.
In the load_from_hub component we do something like this:

dask_df = dd.read_parquet(f"hf://datasets/{self.dataset_name}")

If I see it correctly, the only part that is different here is hf:// vs. gs// or s3://.
What do you think about to adjust the load from hub component to a more generic one?

[mrchtr]

I'd rename it.

Suggested change

	class LoadFromHubComponent(DaskLoadComponent):
	class LoadFromParquet(DaskLoadComponent):

[mrchtr]

Suggested change

	logger.info("Loading dataset from the hub...")
	logger.info("Loading dataset from file...")

[PhilippeMoussalli]

The component does probably the job for now. I am just curious about if we can reuse the load from hub component completely. In the load_from_hub component we do something like this:

dask_df = dd.read_parquet(f"hf://datasets/{self.dataset_name}")

If I see it correctly, the only part that is different here is hf:// vs. gs// or s3://. What do you think about to adjust the load from hub component to a more generic one?

Fairpoint, there is not a lot of difference between them and it can be generalized. I would tackle this in another PR though since it might affect other pipelines (need to change the load from hub component name, examples, docs, ...) #475