Localize setup

[fdrmrc]

Some data structures related to agglomeration had still size equal to the global size of the mesh. This PR removes them.

In particular:

• Removes the concept of "standard" deal.II cells. Every cell is now either a master or a slave (https://github.com/fdrmrc/Polydeal/pull/103/commits/162d458f7499dbaffa9d2f232732b4d0f7c8439f).
• Replace the type of the eulerian vector euler_vector from Vector to L::d::V (LinearAlgebra::distributed::Vector) (https://github.com/fdrmrc/Polydeal/pull/103/commits/4d9f8fd6608b84bf4ff710607eacbfff29216418).
• The vector holding the master-slave relationships, i.e. master_slave_relationships such that master_slave_relationships[cell_index]= -1 if master else master_idx is now distributed too (L::d::V).

The parallel layouts are taken from the distributed tria in the setup phase: https://github.com/fdrmrc/Polydeal/blob/28adfb172f5db935e5a778037d8a59ef9664284d/source/agglomeration_handler.cc#L166-L169

[fdrmrc]

@luca-heltai What do you think? Only relevant commits are 4d9f8fd and 28adfb1

[luca-heltai]

What happens now in the case of standard cells? I remember that in that case we would use the standard FEValues classes. Is this still the case?

[fdrmrc]

What happens now in the case of standard cells? I remember that in that case we would use the standard FEValues classes. Is this still the case?

Yes. Instead of having an enum, we just check if the agglomerate is made by one single cell, like here in the AgglomerationHandler<dim>::reinit(polytope):

https://github.com/fdrmrc/Polydeal/blob/a5b22f1d2874897fd2650b7ac02e20cc9de891aa/source/agglomeration_handler.cc#L497-L498

This was not done in this PR, so such a distinction was in some sense a leftover.