Reimplement open boundary conditions

[JamesMcClung]

Major changes:

• Open particle BCs:
  • Particles that leave through an open boundary are dropped
  • [VB pusher only] Exiting particles deposit extra current
• Open field BCs:
  • Revived the functions that handle magnetic fields near open BCs (I didn't look at what they do, all I know is that reviving them made certain tests pass)
• Fixed several problems with BoundaryInjector

Unfortunately, there are gauss errors at open boundaries. I believe they are false positives caused by inaccurate ghost electric fields. I can think of two possible solutions to this:
a) Communicate ghost corners in the ghost region beyond open BCs
b) Move the particle injection/deletion region to inside the domain
For now, I take solution c), ignore the problem and just deal with the erroneous gauss errors.

Elaborating on the inaccurate ghost electric field:
Consider a particle at the top of a patch, moving left. When it exits the patch, it deposits current in the ghost cells on the left, including some ghost cells in the upper-left corner. The ghost currents in that corner don't get communicated, and thus aren't used to time-evolve the ghost electric fields used by the gauss check, leading to a false positive error. Continuity checks are likely similarly affected.

[germasch]

I found some minor things to comment on ;)