Absolute level of incidence

[kathsherratt]

We previously considered including the absolute level of incidence, as well as the trend.
This was dropped and replaced with a term for the interaction of location x time.
55f0c4a#diff-826594aa74b90636f116b9ab5f64199cbbf5597a35237b758226ea6312fb2323L31

We're now replacing that interaction, with a random effect on dominant variant phase (#138).

Given this, do we want to bring back the incidence smooth term?

• Pros: it complements the idea that we're doing as much as we can to account for the variation across forecast target-specific properties.
• Cons: I don't think it hugely matters either way and possibly better to leave as-is given it's very late in the day for this work

Not keen on changing things around again unnecessarily, so thought I'd raise as a question in draft.