Rework the BoundaryCondition interface

[DanielVandH]

It doesn't really make sense to try and support so many different ways of giving e.g. a Neumann boundary condition - is_neumann_type(type) = type ∈ (:Neumann, :N, :neumann, "Neumann", "N", "neumann") - bit much! I should just define e.g.

abstract type AbstractBoundaryCondition end 
struct Dirichlet <: AbstractBoundaryCondition end 
struct Neumann <: AbstractBoundaryCondition end 
struct DirichletDudt <: AbstractBoundaryCondition end 
# define is_dirichlet, etc. 

and simply go from there. It might even make more sense to force users to instead provide vectors of BoundaryCondition types, e.g. maybe instead of providing functions, types, and params separately, we extend the above so that

abstract type AbstractBoundaryCondition{F, P} end 
struct Dirichlet{F, P} <: AbstractBoundaryCondition{F, P}
     f::F
     parameters::P 
end
struct Neumann{F, P} <: AbstractBoundaryCondition{F, P}
     f::F
     parameters::P 
end
struct DirichletDudt{F, P} <: AbstractBoundaryCondition{F, P}
     f::F
     parameters::P 
end

I don't know if this better for the user experience, though. Maybe the former suggestion is fine.