Support new Dataset API

Bump to arkindex-base-worker==0.3.7rc7.

TODO:

• process_dataset should be renamed to process_set (with most of the code from process_split anyway)
• extract_archive should only do things if the dataset archive was not extracted before
  • if self.dataset_archive is not set
  • or if it's set and the name of its parent folder does not end with the ID of the set's dataset
• We might want to store the cached_dataset as attribute to avoid reloading it across sets
• run should not change that much