Improve dag parsing logic

[PaulFarault]

Which issue(s) this PR fixes

Fixes None

Additional comments

• Isolate the DAG parsing logic from the collection module.
• Collection.dag_nodes now returns a generator.

Agreements

• I hereby declare this contribution to be licensed under the Apache License Version 2.0, January 2004.
• I hereby agree to grant TOSIT a copyright license to use my contributions.

[PaulFarault]

This new version transforms the Collection class to CollectionReader.

The CollectionReader class doesn't store data anymore, only a few metadata remain (collection name, path and paths to directories). This enable to strictly restrict the use of CollectionReader to Collections (that uses it to initialize itself).

To accomplish that:

• I introduced 3 methods that directly read from the collection on call : read_dag_nodes, read_playbooks and read_schemas and deleted the related @property ;

• I added new properties to the Collections class so that it can hold all collections informations (that were previously retrieved from Collection):

  • playbooks a dictionary of available playbooks.
  • default_vars_dirs a dictionary of each collection default var directory path.
  • hostable_entities a mapping of service names to a set of their components.

  This last addition allows to remove the Collections.get_components_from_service method. Collections now only exposes @property.

• Finally, I changed the Collections definition so that it is not a mapping of collection name to Collection instance anymore.

In addition, Collections and CollectionReader has been move to a tdp.core.collections package that only exports the Collections class and a few Error that can be raised. Methods used by the readers methods from the CollectionReader class has been extracted into their own files and unit tests have been refactored accordingly.