Setting Dirichlet boundary conditioners in preconditioners

[maddyscientist]

The present approach to setting zero boundary conditions in Schwarz solvers and enabling / disabling global reductions for these solvers is haphazard and limited. We need to expose a cleaner method for doing this.

Perhaps the solvers should query the Dirac operator in question what its boundary conditions are, and if it has Dirichlet boundary conditions we can disable global reductions?

Anyway, this is something we should clean up after #903 is merged in.

[cpviolator]

@maddyscientist How are Dirichlet BC presently enabled applied? I'll need to implement this more generally in the near future.