sd-boot: simplify the implementation of entry uniquification

sd-boot: simplify the implementation of entry uniquification

There's a slight change in implementation: we first try to append the
version, then look for any non-unique pairs again. Before, we would only
mark as possibly unique those entries we changed. But if there are two
entries that e.g. have the same title and version, but only one has the
machine-id specified, we would treat one of them as still non-unique after
appending the machine-id to the other one. So the new algorithm is simpler
but more robust (not that it matters).