1) Ensure that Last-Translator is always set when pulling from
Transifex. If there are no reviewed translations, Transifex will
not include that header, but our validate code requires it.
2) Don't complain about obviously missing files when cleaning
the headers of just-pulled files, and make sure that we do
notice the files we did actually pull.
3) Ensure that Plural-Forms has a valid value when extracting.
python-gettext will choke otherwise.
4) When creating a dummy po file, don't garble source strings in
a way that relies on the Python dictionary order. This way the
result will always be the same no matter the machine.