Incorporate uncertainty propagation between LDA and change point

[juniperlsimonis]

currently, the best fit (by AIC) LDA is all that's passed to the change point model. we should consider propagating the full posterior distribution of the LDA through the change point model. in certain situations, the LDA may be uncertain enough that it impacts the change point model, and we want to be able to tease that apart.

[juniperlsimonis]

this should be accomplishable with #149 and incorporation of replicated runs of a given LDA_TS (perhaps replicates to show variability or variations of a cross validation schema)