bugfix: gendummydata.py was producing invalid usernames

As per our regex and policies, usernames should consist of
ascii alphanumeric characters and possibly (-, _ or .).

gendummydata.py was creating unicode versions of some
usernames and adding them into the DB. With our newfound
collations, this becomes a problem as it treats them as
the same.

This should have never been the case here, and so,
gendummydata.py has been patched to normalize all of its
usernames and package names.

Signed-off-by: Kevin Morris <kevr@0cost.org>