Interaction kappa1, kappa2 and S in TN93

[richelbilderbeek]

Feedback bjoelle:

creating a TN93 model with the default settings creates two log-normal distributions as priors for kappa1 and kappa2 where the S parameter has lower=0, upper=0 and value=1.25. This works if S is not estimated, because BEAST2 only checks upper and lower when modifying the value (not when initialising it) but it will break if S is estimated.