#527 Replace the alias GlobalVector and GlobalMatrix by a struct inherited from the Petsc Wrappers. The goal will be now to enrigh the newly introduced classes.

#525 Get rid of the clumsy way that attempted (and yet didn't succeed completely) to choose which pattrens to build. Now all combinations are built, and it eases considerably the creation of a model. Avoiding to build unused patterns is the topic of #520.

#525 Remove numbering subset storage from base variational formulation, and the clumsy selection of what matrix should be built. The derived class is much better placed to know what is actually required or not and should therefore be in charge. Used only with monolithic Stokes so far.

#525 Stokes: numbering subsets are called by a different way, which is more appropriate to the genericity I want to achieve.

#517 Elasticity: remove the temporary input parameters introduced to highlight and correct the initial bug.

#521 NonMonolithicStokes: Begin to introduce new formulation in the model. Add the missing method definitions.

#521 NonMonolithicStokes: implement two functions to compute the new velocity and pressure, respectively.

#521 NonMonolithicStokes: Enrich new VariationalFormulation with attributes from both existing ones.

#512 VariationalFormulation: extend the allocation to truly cover cases with several numbering subsets.

#512 Refactoring of VariationalFormulation in progress. There is now the interface to store multiple global matrices; for global vectors a dummy one is there (the calls use the numbering subsets but the storage still keeps only one; it will be added in the next commits). Only Elastic model is up-to-date and is currently working.