Unify code path of DirichletBC::set

[schnellerhase]

Reduces code duplicity in the set functionality of the Dirichlet boundary conditions.

This also highlights, that in general the fundamental operation we need to implement to set a Dirichlet boundary condition is the one provided by the set_fn. So as long as one can define a mapping from index of Dirichlet boundary condition DOF to a coefficient value the DirichletBC::set can be defined. Meaning in the future the DircihletBC class or set function may be altered to take a general set_fn instead of a specific type that contains the values - this would follow the functional style employed by the assembly routines and allow easy extension and remove the type dispatching within the DirichletBC class.

[garth-wells]

Looks good. Just needs some small clarifications and simplifications.