Implementation: Scaling of G specific Components of Substitution Matrices

[mikegilchrist]

This seems like our highest priority.

We need to determine if the issues with the Random Effects for Sensitivity model come from using different time scales with different partitions, i.e. G values.

If this is the case, the fix should hopefully be easy and will resolve the problems.