AMR optimization

[landinjm]

Moved the refinement manager to a source file.

Still working on some refactoring and cleanup.

Seems like part of the reason for #1082 was the construction of the multigrid hierarchy. We shouldn't do that if we don't need too.

[landinjm]

Most of it had to do with the FEValues being constructed in the loop. Just like the issue with nucleation.

I'm planning to move these to SystemWide where they can be lazily initialized.

@fractalsbyx FYI

[landinjm]

I think we can get away with a single FEValues for scalar and vectors field that works for our modules (AMR, nucleation, integration, grain remapping)

  static dealii::FEValues<dim, dim> &
  vector_fe_values()
  {
    static dealii::FEValues<dim, dim> fe_values(
      mapping,
      fe_systems[1],
      quadrature,
      dealii::UpdateFlags::update_values | dealii::UpdateFlags::update_gradients |
        dealii::UpdateFlags::update_quadrature_points |
        dealii::UpdateFlags::update_JxW_values);
    return fe_values;
  }