fold secondary options into observation model

[sbfnk]

secondary_opts has just one named option: type ("incidence" or "prevalence"). This could be added to the observation model options instead, and a calculate_secondary call be added to the estimate_infections model (making it possible to fit it to incidence and prevalence data).

There is also the option of modifying any of the lower level parameters (cumulative, historic, primary_hist_additive, current and primary_current_additive) but in absence of a use case (which I may just not be aware of) we could deprecate that option.

[sbfnk]

This would be a preliminary step to enabling https://github.com/epiforecasts/EpiNow2/discussions/371