Conditional coalescent prior

[awohns]

We currently estimate the location and scale parameters from the grid_data field of the tsdate. NodeGridValues priors object. This is a bit strange, since tsdate finds the expectation and variance of the age of each node, then uses method of moments to fit a log normal prior distribution, then discretizes probability into timepoints bins. Instead of using the discretized times to estimate the log normal, we should just use the actual location and scale parameter. These don't appear to be saved in the output from tsdate, so I'll need to make the change over there first.

This likely won't make much of a difference, but it's definitely more correct.