Review and refactor prior handling

[seabbs]

#474 has made the prior interface somewhat complex in response to a range of edge cases found during that PR (in particular manual priors used for the event windows). It may be possible to simply this interface or at least to reduce to complexity exposed to the user.

This issue is probably blocked by #221 as that model may have other prior needs that factor into this refactor.

[athowes]

Agree this makes sense to tackle after adding the marginal model (#221).