Predict to any specified age brackets

[njtierney]

Some notes from Golding:

To create the bins we need to integrate the contacts over the age group. We currently do that with a finite sum over a 1 year time step, we just need to do that on a finer grid of timesteps, based on the age breaks provided

[njtierney]

This would be the relevant section of code to change: https://github.com/njtierney/conmat/blob/d694262b09de15a9f8cd48a47fabf074f94e517d/R/predict_contacts.R#L87-L100

[njtierney]

Assigning to @MikeLydeamore

[njtierney]

Started in #108 - this is on the right track, we need to essentially avoid using age_min:age_max to create the age range we will predict over. This PR uses a more flexible approach to manage how the years are constructed. And it should probably have an option for the user to pass through some default length.out (e.g., it could be 1000 by default).

Check how the aggregate function works with this - it should be OK but worth checking