Apply g-tensor to spin dipoles before calculating dynamical correlations

[kbarros]

Sunny is not currently consistent about whether the dynamical correlations involve the angular momentum dipole, or the magnetic moment dipole. The difference becomes nontrivial when the $g$-tensor is not a scalar. Previously, the $g$-tensor was applied by default only in :dipole mode, and only when using the classical dynamics to estimate intensities. We should unify the behavior across all modes (:dipole and :SUN) and across all calculation paths (spin wave theory and classical dynamics). In all cases, the $g$-tensor should be applied by default, to make contact with the neutron scattering cross section. To get spin-spin correlations instead, there should remain the option apply_g = false in the construction of an intensity_formula.