Improve our handling of file encodings with diffs.
The way that we dealt with diff and file contents was kind of broken with

regards to special encodings. This was made worse by the unicode transition,

and in particular, I noticed that the diff validation resource would raise an

exception if the diff had unicode characters in it. Fixing that inspired me to

fix all our issues.
We now decode strings to unicode before manipulating them, and then encode them

back to bytes afterward using the encoding that we used to decode. This means

that we're no longer splitting lines or doing regex replacements on bytes, only

on unicode objects.

Ran unit tests
Uploaded a diff containing unicode text via the "New Review Request" form.