Normalization of the power spectrum

[abensonca]

Normalization of the power spectrum is currently handled in a non-optimal way. We specify a sigma_8, the cosmologicalMassVariance class is then responsible for normalizing to match that, and provides a utility method that gives the resulting factor by which the power spectrum should be multiplied to get the correct normalization.

We also have an option to match the power at some reference wavenumber in another power spectrum instead.

It would be better to have a dedicated powerSpectrumNormalizer class which provides implementations corresponding to each of these methods (and, potentially, others). This would need to be called by the powerSpectrumPrimordialTransferred class to normalize itself (so would need to live in the same module I think).