The way the RBTools Python API works, it's annoyingly easy to make GET

requests to list resources when we don't actually care about the

existing data, we just want an instance of the list resource in order to

to do POST operations.
This change fixes rbt post to minimize this in two big cases:

- We were fetching the review request list when creating a new review

  request. We had only-fields= on this request, so it wasn't actually

  serializing very much data, but it would still query and return a list

  of up to 25 review requests.

- We were fetching up to 25 diff file attachments when we wanted to

  upload binary files, including all of their data. Moreover, we were

  doing this for each binary file we wanted to upload.
With this fix, we now pass max_results=0 to the get operation. This

means we'll get the proper MIME types and links which are used to do

further operations, but we won't fetch any existing data from the

database.
The file attachments resource is now saved once instead of fetching it

for each file.
This also fixes up the DiffFileAttachmentListResource upload method to

use the create link (before it would inherit the ?max-results=0 from

the GET).

Posted changes and looked carefully at the --debug output. Verified that

we were no longer doing unnecessary or unnecessarily heavy API requests.