Latent event times as distributional parameters?

[athowes]

In previous code versions the latent parameters were set to be distributional rather than injected with the Stan code. See here.

This has the benefit of allowing users to set priors on them with brms interface (#199). Users could then e.g. set priors on the primary event time to incorperate epidemic growth, or on either event time to incorperate e.g. daily cycles (event less likely at night).

The scope for this issue is to investigate whether it would be possible to refactor what we have to use this.