* For data projects:
  - create() to create datasets
  - setindex!() to add existing datasets
  - delete() to delete datasets
  - Implementations for StackedDataProject, AbstractTOMLDataProject and TOMLDataProject

* Concrete save_project() API to persist a DataProject to a file as TOML

* For storage drivers:
  - AbstractDataDriver and implementation for FileSystemDriver
  - open_dataset to do what the current function-based API does
  - create_storage to initialize storage
  - delete_storage to remove storage
  - These ideas seem a bit half-baked

* Refactoring open() to add write=true keyword

Fix save_project for delete and create