feat(core): new dataset provenance

[m-alisafaee]

Description

Adds two indexes to the new persistent layer: datasets and datasets-provenance. The first index is a mapping from dataset names to data objects for all current datasets in a project. The second index is a mapping from dataset id to dataset object; this index contains current datasets and all previous versions of all datasets.

Fixes #2119

[Panaetius]

Would it be cleaner to have two methods, add_or_update_dataset_provenance and remove_dataset_provenance?

It would also be cleaner to just call the methods on DatasetProvenance directly and inject that where needed, since this method in LocalClient doesn't really contain any additional logic (especially since commit_database isn't set to True anywhere). In the future I'd like to rely on LocalClient less and have the dedicated classes do more of the heavy lifting.
If we inject DatasetProvenance using the binder.bind_to_constructor(Class, lambda: Class(*args)) way of injection, it would only initialise DatasetProvenance if code that uses it (well, requests it as an injection) is called and then it would be a singleton, so we could just inject this, ProvenanceGraph and DependencyGraph this way in the with_database CommandBuilder without performance overhead, and it'd save performance since we wouldn't have to do from_database more than once.

[Panaetius]

I know I wrote this, but now I'm confused, does this line actually do anything other than raise an uncaught error if the redirect url is not a valid url? I feel like we could remove this line? Looks like a copy&paste error by me