queryset.sum and queryset.average should use the aggregation framework

[wojcikstefan]

Using map-reduce is slower, uses both locks (r & w - see http://docs.mongodb.org/manual/faq/concurrency/#which-operations-lock-the-database) and isn't necessary.

@rozza take a look at closeio@50d43c9 and let me know if you like the solution. It works for us and is much faster:

In [4]: %timeit Activity.objects.filter(organization=org).sum('duration')
1 loops, best of 3: 1 s per loop

In [5]: %timeit Activity.objects.filter(organization=org).aggregate_sum('duration')
10 loops, best of 3: 171 ms per loop

In [6]: %timeit Activity.objects.filter(organization=org).sum('duration')
1 loops, best of 3: 989 ms per loop

[wojcikstefan]

@MRigal does this look interesting to y'all?

[MRigal]

This makes fully sense and is more than interesting but almost required to me !!! I have a couple (dozens?) of old bugs and issues I want to wipe first, but it is one of the thing I would be very happy to do in the near future !

If you're up for a PR to implement the Aggregation Framework, please do so ! You may start with reusing the commit, but you may also do more, by looking at some proposals/discussions in the other issues labeled "Aggregation Framework"