Change return type of File Converters to List[Document]

[bogdankostic]

Currently, the file converters are returning List[Dict]. We should make use of our Document primitive here and return List[Document] instead.

[tholor]

Same refactoring is needed for the other nodes in our indexing pipeline (e.g. Preprocessor).

When working on this, we should take into account that we create the document's id at the moment when we instantiate the Document. If this now happens within different nodes, we probably need to add a param there to control the way how these ids are created (default: hash of content). For example, in DC we want to create the ID based on content + a pipeline_id. This can be done via id_hash_keys parameter in Document.init().

Ping @ArzelaAscoIi when kicking the work off here.