Hash collision with renames, blank files, etc.

[siddhesh]

The hash computation algorithm does not take into account file renames or new file creations, because of which if a diff has only renames or a blank file either being created or deleted, they all end up having the same hash. At the minimum, the filenames (and the operation on it) being hashed should be added to the hasher to eliminate these conflicts.

For example, these two patches end up with the same has with hasher.py:

https://patchwork.sourceware.org/project/glibc/patch/7c1fb956621e5644de7342022e4ae599b79b5626.1613577607.git.fweimer@redhat.com/
https://patchwork.sourceware.org/project/glibc/patch/20140621021718.2A4AA2C39AB@topped-with-meat.com/