Fix non-constant effective mass

[viljarjf]

Change the current nabla^2 to nabla(m*nabla) to fix

[henrtk]

Is it sufficient to add $\nabla m\cdot \nabla$ term by direct matrix multiplication, ie. $\nabla\cdot(m\nabla \Psi) = \nabla m\nabla \Psi + m\nabla^2 \Psi$?

[viljarjf]

Continuity at mass boundaries is not satisfied with that approach. See test_array_mass in tests/test_hamiltonian.py