Improve statistics' gathering mechanism

Analyze Zanata's statistics' collection mechanism and improve it. A few points to keep in mind:

• Statistics should be quick to fetch

• Keep statistics as up-to-date as possible

• In some cases (e.g. large untranslated projects) calculating statistics takes a long time.

Implementation options:

• A backend cache is currently used to store calculated stats. This approach seems appropriate, but the entries are recalculated every time there is a change. There are cheaper ways to do this.