Per-user quotas and quota data model refactoring

Considering the notorious story if the original per-user quota blueprint (implemented, committed, then rolled back) it becomes apparent that introduction of another dimension of resource limitation (i.e. user-tied quotas and, maybe, quota classes) is almost impossible without bringing in significant changes to the internal resource accounting subsystem.

The goal of this blueprint is to propose a new data model for the quota subsystem which could make the user-based quotas possible.

The following items should be covered in the specification
- the description of the 2-dimensional (user & project) quota lookup and resource reservation algorithm
- the detailed description of the entity-relationship model this algorithm could rely on
- the re-consideration of the data backend effectiveness in the context of the new model
- the preliminary migration plan for already existing installations of OpenStack