The Poisson equation was not working because I didn't finish
implementing the new type of BCs. Dirichlet is probably untested. I
should stop doing shitty developments and no testing.
I should've commited before, but I wanted to make things compile.
The big change is that I've added a global time step so the parameter
does not need to be passed in each function. This is useful as we are
moving towards using time profiles for boundary conditions and injection
of particles (not in this branch, but in the future and the procedure
will be quite similar)
