When we make a dataset we don't know if it's valid and able to be submitted for evaluation until we click the submit button. In order to give earlier feedback we could do some basic checks in the dataset editor. These checks should only happen in edit mode, or if the dataset is owned by the current user

• On the main dataset view page, given an indication if a dataset doesn't have at least two classes
• On the class page, give an indication if a class doesn't have at least two recordings

These minimum values should reuse existing dataset constants: https://github.com/metabrainz/acousticbrainz-server/blob/f9afa4017c49b7d412b6557b68744d8fbbb2da07/db/dataset_eval.py#L109-L110

In AB-148 we want to allow datasets to be submitted even if they contain MBIDs that don't exist in AB. We can also add this check to the dataset when editing it. For example, when retrieving metdata for the recording to show on the dataset editor page we can also see if this recording is in AcousticBrainz: https://github.com/metabrainz/acousticbrainz-server/blob/b064542deb390483e81795619b911c97dd8dd606/webserver/views/datasets.py#L398

With this information retrieved in the dataset editor, we can also show a message saying that even though there are more than 2 recordings, the class is invalid because some will be removed.