[Datasets] Remove FastFileMetadataProvider warning

[bveeramani]

Why are these changes needed?

If you use a datasource that uses FastFileMetadataProvider (e.g., ImageFolderDatasource), a warning is raised:

(_prepare_read pid=43380) 2022-05-11 02:11:33,272 WARNING file_meta_provider.py:199 -- Skipping expansion of 2 path(s). If your paths contain directories or if file size collection is required, try rerunning this read with meta_provider=DefaultFileMetadataProvider().

This may confuse users because:

• It has nothing to do with the user's problem
• It leaks an implementation detail
• It can't be resolved (there's no issue in the first place)

Related issue number

See #24641

Checks

• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[jianoaix]

Ah, I think I commented this before that it should probably not log unconditionally.

[ericl]

Why not have ImageFolderDatasource use the default metadata provider instead? The warning is here for a reason: not resolving sub directories is very dangerous behavior.

[bveeramani]

Why not have ImageFolderDatasource use the default metadata provider instead? The warning is here for a reason: not resolving sub directories is very dangerous behavior.

The current implementation of ImageFolderDatasource is too slow with the default metadata provider.

[bveeramani]

Once #24822 gets merged, I'll use the default metadata provider to implement ImageFolderDatasource.

That being said, I agree with @jianoaix that we probably shouldn't unconditionally log this warning. I don't know if it makes sense to log a warning when FastFileMetadataProvider is being used correctly.