[13/14][Adjoint Module] Cover mode monitors under a mirror symmetry

[smartalecH]

test_adjoint_symmetry.py (added earlier in this stack) exercises FourierFields only. EigenmodeCoefficient reaches the adjoint through a different path, and the gap is not academic — it let a wrong mirror eigenvalue in the example at the top of this stack look like a solver fault for some time.

What a mode monitor adds is that its mirror eigenvalue is not always +1. A mirror in y maps y -> -y and flips the sign of y-components, so a mode whose Ey is even in y has eigenvalue −1. Declaring the wrong one does not fail; it projects onto the orthogonal sector and quietly returns a different answer. In the 3D grating coupler that showed up as a mode coefficient 1.51x the unsymmetrised one and a design gradient 16.8x too large against a converged finite difference.

Here Ez is even about the plane, so +1 is the right eigenvalue, and the tests assert what must follow:

• the objective is unchanged by declaring the symmetry;
• the gradient matches a finite difference under it;
• the folded gradient matches the unsymmetrised one.

The fold is the part worth stating. The raw gradient is not invariant: Meep evaluates epsilon only on y >= 0, so weights below the plane are inert and one above it drives both halves, and the raw result comes back doubled above the plane and zero below. Folding the halves together recovers the derivative with respect to a variable that owns both mirrored cells. Getting that wrong is what makes the correct behaviour look like a bug.