Speed up disk quota computation

[AndrewJGaut]

Reasons for making this change

Current method of updating disk quota is slow. Basically, currently the disk quota is updated using the following function:

def _get_disk_used(self, user_id):
        return (
            self.search_bundles(user_id, ['size=.sum', 'owner_id=' + user_id, 'data_hash=%'])[
                'result'
            ]
            or 0
        )

and then setting the disk quota to the returned value. However, doing such a search is quite slow, and so we want to do it faster. This PR speeds up the computation and just increments the user's disk quota each time by the correct amount.

This PR also adds a bunch of disk quota tests.

Related issues

#4350 and #4351

[AndrewJGaut]

TODO @AndrewJGaut : Add benchmarking

[epicfaace]

why remove this?

[AndrewJGaut]

You can't see it in this abridged view, but if you look at the full file the code looks ilke:

 try:
                ....
                self._client.update_bundle_state(bundle['id'], params={'success': True})
            except Exception as err:
                ...
            else:
                self._client.update_bundle_state(bundle['id'], params={'success': True})

What does this mean? Well, the 'else' branch triggers in a try-catch block when no error is thrown. In other words, when no errors was thrown, the self._client.update_bundle_state(bundle['id'], params={'success': True}) call was triggered twice. This wasn't an issue (except for efficiency reasons) before since the update wasn't incremental (it just reset the whole value), but it was for me since the disk increment happened twice.

[epicfaace]

Do these test code paths end up calling cleanup_existing_contents?

[AndrewJGaut]

Ah, shoot. No, I don't think so. Let me add one more test for that before I merge.

[AndrewJGaut]

Thanks for catching that!