#1545 See if we really need to store the complete matrix pattern
We saw in #1526 (closed) that the geometric elements related to boundary conditions were not properly kept during the reduction to processor-wise process.
My hunch is that the missing dofs in the matrix pattern if we try a recomputation from the prepartitioned data (or simply if we attempt to compute the pattern after the call to GodOfDof::ReduceToProcessorWise()
) share the same cause; so I will try to keep explicitly the boundary conditions dofs.
I won't spend too much time here, but if I'm right it might make the GodOfDof initialization leaner and help with the following step (do the reduction during a sequential run).