Update RCB partitioner for cartesian mesh (through Zoltan)

[DavidDureau]

Context:

• the whole domain is a cartesian mesh with cartesian patches;
• the cell edges are parallel to the XYZ axis;
• the number of ghost layers is greater than or equal to 1;

Goal: to have a RCB initial partitioning (thanks to Zoltan?)
The partitioning is static along the simulation.

[grospelliergilles]

Duplicate of #128

[DavidDureau]

The case where the cost per cell is the same for all cells is interesting.
Now, the number of partitions is equal to Ni x Nj x Nk which limits the number of MPI processes and can give partition with a bad aspect ratio.

Thanks to the RCB partitioner we would have the whole domain decomposed into P partitions, whatever P.

Could a code have a default RCB partitioning considering a uniform cost for all cells?
(in this case, the code needn't define a cost per cell)