Unify estimation across years

[curryc2]

Currently the run reconstruction model estimates each year semi-independently. By this I mean estimation is separate for each year, but some information is shared among years (availability).

It would make sense to treat the run reconstruction in a fully hierarchical fashion, estimating each year conditional on values for key parameters in other years.

Selectivity parameters for example could be assumed to arise from a normal distribution with a group mean and standard deviation, suggesting (as was shown in Cunningham et al. 2017) that selectivity generally increases with ocean age, while also providing for inter-annual variation in selectivity values.

This likely requires treating availability deviations as a random effect and may necessitate re-coding in TMB.