gen_testdata.py picks up root keys randomly and shuffles extension dict
keys also randomly. But even with predictable PRNG if the input to e.g.
rand.choice is different, the result will be different as well.

So far we were lucky: we were running gen_testdata.py only via py2 and
the order of retrieved root keys was - by chance - the same each time.
That's why generated testdata databases were the same after each
gen_testdata.py run. But even on py2 there is no such guaranty and when
runnning gen_testdata.py via py3 the order of keys is really different:

    $ python2
    Python 2.7.18 (default, Jul 14 2021, 08:11:37)
    [GCC 10.2.1 20210110] on linux2
    Type "help", "copyright", "credits" or "license" for more information.
    >>> d = {}
    >>> d['a'] = 1
    >>> d['b'] = 2
    >>> d['c'] = 3
    >>> d.keys()
    ['a', 'c', 'b']

    $ python3
    Python 3.11.2 (main, Mar 13 2023, 12:18:29) [GCC 12.2.0] on linux
    Type "help", "copyright", "credits" or "license" for more information.
    >>> d = {}
    >>> d['a'] = 1
    >>> d['b'] = 2
    >>> d['c'] = 3
    >>> list(d.keys())
    ['a', 'b', 'c']

So let's prepare for py3 generation beforehand by making sure that
keys input to PRNG is the same be it py2 or py3, thus, giving a chance
for generated py2/py3 databases to be really close to each other.

Since here we change the order of keys that are feed to PRNG, generated
test databases are shuffled a bit.