BulkDeleter: normalize return keys and support > 10000 objects

[sivel]

This pull request brings back functionality that was removed in 1.9.

1. normalize keys to match the documentation (Ref: bulk_delete() returns a dictionary with wrong keys #479)
2. Support deleting more than 10000 objects (Ref: container.delete_all_objects() doesn't function properly #392)
3. When we ask to delete all objects, actually list them all (Ref: container.delete_all_objects() doesn't function properly #392)

[bazzilic]

Having looked through your PR, I have a suggestion: in case Response Status is not 200 OK map Response Body to errors.

[sivel]

errors should already be populated in such a case. If I add additional information into errors the length of the list will not match the actual number of errors:

http://docs.rackspace.com/files/api/v1/cf-devguide/content/Bulk_Delete-d1e2338.html.html#d6e8440

I'm still working on fixing some related issues, so there will be more commits added soon. Specifically, I am also working on some issue that are part of #392

[bazzilic]

I had a situation today, which led to me opening this issue. I was trying to delete 12k objects, CloudFiles returned a status 413 Request Entity Too Large. Request Body contained an explanation that we can only delete no more than 10k objects in one request. However, the Errors list was empty.

[sivel]

Ok, I'll try to look into that. Part of this PR is also to fix the 413 Request Entity Too Large issue.

[bazzilic]

Yes, I noticed that. I am doing it a bit differently, but your approach looks more memory efficient.

page_size = 5000
for batch in [names[i:i+page_size] for i in range(0, len(names), page_size)]:
    ctx.get_client("object_store", region, public=False).bulk_delete(container, batch)