This PR just provides a parallel version of the classical DG for the Poisson problem without using the MeshWorker interface. The main reason for this is that with polygons we are forced to assemble jumps and averages of the DG formulation "by hand".
Ideally, the parallel version with polyDG will look almost identical to this (except of course for the objects needed by the polygonal infrastructure). @luca-heltai Do you agree, or do you think we should aim to a different interface? (Here I'm not strictly talking about our internals, which I am going to hide/refactor according to #68, but only from a user's perspective)
This PR just provides a parallel version of the classical DG for the Poisson problem without using the
MeshWorkerinterface. The main reason for this is that with polygons we are forced to assemble jumps and averages of the DG formulation "by hand".Ideally, the parallel version with polyDG will look almost identical to this (except of course for the objects needed by the polygonal infrastructure). @luca-heltai Do you agree, or do you think we should aim to a different interface? (Here I'm not strictly talking about our internals, which I am going to hide/refactor according to #68, but only from a user's perspective)