Normalizations

Need to make sure all distributions are normalized (or normalizable numerically in an efficient way). This can probably be split into "logpdf" and "lognorm" function calls, but perhaps it doesn't need to be.

The issue is that if we ever want to condition one distribution on another (it evolve parameters with another variate) then we need to be able to normalize the distributions in order to compute the likelihoods we need (marginalization a over single-event parameters).

Identified in discussions with @maya.fishbach