Pull out diff-related data into a "diff context" object and initialize models from it.
The diff viewer needs a bunch of bits of data, which we render in a django

template as info in a <script> tag and then consume from javascript. This

change centralizes all of that info into a "diff_context" object which is

rendered using the json_dumps filter tag. This object is then consumed by a new

DiffViewerPageModel, which will parse the high-level object and create/update a

bunch of child models, which are then consumed by the various views that make

up the diff viewer UI.
Coming soon, we'll be able to consume this diff context either from the

javascript rendered into the template, as well as when it comes from an API

payload.

Tested various pieces of the diff viewer UI to make sure that the initial

  data was still correct.
Used this in conjunction with other changes that load different parts of the

  diff viewer without reloads.
Ran unit tests.