Perforce was doing a bit too much work when checking for the existence

of certain files and when parsing files at revision 1. For file

existence checking, we were using the default behavior of fetching the

complete contents of the file and using that to determine existence.

That wasn't the most optimal, especially for very large files. We now

use p4 fstat to look up the file instead.
For revision parsing, we had to work around some behavior in older

versions of Perforce where new files had a revision of 1 in the diff. To

determine whether the file existed in the repository, we ran p4 files

and checked the result. However, we weren't limiting the number of

results, and so files with a lot of history were slow. Also, for newer

versions of Perforce, revisions for new files are 0.
Instead of running p4 files, we just use the new file existence

checks. This has the advantage of using/populating the existence cache,

helping reduce the number of remote calls even further. We also now only

run these tests if the revision is 1 (and we assume 0 means it's a new

file).

Unit tests pass.
Tested these with a real diff against a real server consisting of new

files at revision 0 and with files with large files containing lots of

history. Saw the results I expected, and reduced times needed for these

functions.