Add support for randomized ordering of fakequeryset

[KIRA009]

Fixes #111

This PR randomizes the queryset by using random.random() values as the key during sorting of the results.

[KIRA009]

I am wondering how do we test this behaviour?

I thought of something like this, where we run the query n number of times, and if the ordering is different from what was seen previously, the test passes.

recognized_order = ' '.join(tuple(album.name for album in beatles.albums.order_by('?')))
for _ in range(100):
    new_order = ' '.join(tuple(album.name for album in beatles.albums.order_by('?')))
    if new_order != recognized_order:
        break
else:
    self.fail("Ordering was the same in all 100 queries")

But I am not sure if this is the best way to do this

[gasman]

It's probably fine to just run it once and verify that it contains the expected items - since that's the only thing you can rely on when using it in the real world.

[gasman]

I think it would be better to use random.shuffle in this case, and avoid going through items.sort / get_sort_value entirely:

if key == "?":  # Random ordering
    random.shuffle(items)
else:
    def get_sort_value(item):
        # ....
    items.sort(key=get_sort_value, reverse=reverse)

[gasman]

Merged in df4c6be - thanks @KIRA009!