[FEATURE] Make LoadHubDataset more general to read local files

[plaguss]

Is your feature request related to a problem? Please describe.
I created a custom step to read data from a jsonlines file in the local fileystem, but noted that load_dataset already has the functionality, we only need to expose it.

Describe the solution you'd like
Allow reading local files using LoadHubDataset, something similar to the following

load_dataset = LoadHubDataset(
    filetype="json",
    filename="path/to/dataset.jsonl"
)

Describe alternatives you've considered
Creating a custom step when needed.

Additional context
Ref: https://huggingface.co/docs/datasets/loading#local-and-remote-files

[alvarobartt]

I also wanted to bring up the fact that we now have LoadDataFromDicts and LoadHubDataset; most of the times, myself, as an user, end up writing LoadDataFromHub assuming the standard is LoadDataFrom ...; maybe this is a nice time to unify everything under the same naming pattern? I believe this will be easier and avoid conflicts in the future (we can set a deprecation warning on LoadHubDataset for the next couple of releases to avoid breaking any existing pipeline).

[rasdani]

After the merge of #673 there also needs to be a LoadS3Dataset or not?

Like this:
#665 (comment)

[plaguss]

Closing this with #691. @rasdani will let you know once we add the docs, but the you could use the new LoadDataFromDisk to work with the s3 datasets saved to disk.