implement Dataset.load(source, LoaderConfig, **kwargs)

[johentsch]

Like for features, LoaderConfig can be either the name of a loader, a Loader object, or a Config describing one. If not specified, dimcat will infer the suited loader in this order:

• If the source contains datapackage.json files, use PackageLoader
• If the source contains resource.json files, use ResourceLoader
• If the source contains metadata.tsv or concatenated_metadata.tsv (i.e., likely a DCML corpus or otherwise parsed using ms3) use MuseScoreLoader
• else, decide based on the detected file extensions