Re: Honoring a checked out gitattributes file

Ok, here's another round. I fixed the test case to expect failure, as
pointed out by Matthieu, and I also added code to make sure that
.gitattributes gets checked out first.

I also added a test case and some code to support the case where
.gitattributes is removed in a commit, but this doesn't work properly
yet (see commit message). I'm not sure how to solve this without
resolving to ugly hacks like passing the new index_state* all the way
down to where git_checkattr gets called. If anybody has any
suggestions, do share.

The main usecase where .gitattributes is modified, is anyway covered
by this patch.

--
Kristian