Using git version is simpler as it handles oddball cases internally.
Since we have been using mostly lightweight tags I have set it to
search those in addition to annotated tags for the best match. The
output for commits that don't have a tag seems a bit nicer as well.
If there are multiple tags pointing to HEAD we need to pick only
one. We'll take one that seems to be a version number. This allows
the version number tag to coexist with another tag, that perhaps
indicates which images it was used to build.
The previous system required making commits in order to do a new
build. This caused problems because we were using the git repo
directly for fedora builds and we couldn't use that exact version
when building matching packages because of needing to make new
commits in order to do the build.
Having the spec file in here also was unnecessary. It is easy to
just maintain the spec file in the Fedora package.
I still want to do a bit more future work with this to pick up
the version from a git tag.
