fix: allow for both linear and quadratic dissipation functions with SpectralSplit

[CusiniM]

There was an hardcoded preprocessor variiable that forced the computation of the degradation function for the SpectralSplit to only be valid for a linear dissipation function. This PR removes that ifdef and makes both options available.

Because of the way the model is currently used I had to come up with a mixed solution. The selection is done statically for the phase field damage kernels but dynamically for the smallStrainUpdate. It's fairly easy to template the full damage model but I want to have a chat with @frankfeifan about the formulation and what degradation/dissipation functions combinations are actually valid.