Refactor Celery and Analysis Handling

[ilan-gold]

The code currently for handling the end of an Analysis is slow and redundant. It is requires rethinking how we handle the output from galaxy and how we create objects.

The work has already begun with fixing #3435 and #2642, but requires more, namely

1. removing/refactoring redundant code
2. refactoring tests (a big lift considering the amount of mocking going on)
3. potentially altering the way in which we handle the tasks at the end
4. using Node and not AnalysisResult to handle the post-analysis files - this will help resolve Analyses fail randomly due to IndexError #2993
5. Add progress tracking for other stages beyond just the actual Analysis