Add new CurlCurlSolve() to propagators

[Monoclod]

Add a new propagator called CurlCurlSolve() to find weak solutions to the curl-curl problem.

Find $\mathbf{E} \in H(\textnormal{curl},\Omega)$ such that

$$ \int_\Omega (\nabla \times \mathbf F) \cdot (\nabla \times \mathbf E)\ \textrm{d} \mathbf x - \sigma \int_\Omega \mathbf F \cdot \mathbf E\ \textrm d \mathbf x = \sum_i \int_\Omega \mathbf F \cdot \mathbf J _i\ \textrm d \mathbf x \qquad \forall \mathbf F \in H(\textnormal{curl}), $$

where $\mathbf J _i:\Omega \to \mathbb R^3$ is a real-valued vector field and $\sigma \in \mathbb R$ is a scalar.
Boundary terms from integration by parts are assumed to vanish.

The equation is discretized as

$$ \left( \mathbb C^\top \cdot \mathbb M^2 \cdot \mathbb C - \sigma \mathbb M^1 \right)\cdot \boldsymbol e^{n+1} =\sum_i \mathbb P^1 \cdot \boldsymbol J _i , $$

where $M^1$ and $M^2$ are objects of the struphy.feec.mass.WeightedMassOperators class, and $\mathbb P^1$ is the projector into the space $V^1_h$.

The method does not converge properly for sufficiently high spline degrees and sufficiently fine mesh grids due to the inherent ill-conditionedness of the curl-curl problem; this should be solved by proper initial conditioning.

[spossann]

@Monoclod well done! Please add some tests that prove that the new propagator works. For the convergence of higher degrees, I suggest you take a function with more variations, so that the error is larger, and set tol=1e-7 in the solver.

[Monoclod]

Added a test for the propagator (and fixed a few comments). They seem to converge fine even for tolerances of 1e-10, oddly better than my previous attempt

[spossann]

@Monoclod can you merge with devel to see if the tests pass?

[spossann]

will be merged from another branch.