Allow fixed background hazard by covariates

[chjackson]

Background mortality is commonly reported by covariates, generally age, sex and calendar year. The cetuximab example data from Guyot et al. uses a fixed pre-processed background hazard that represents an "average" over people like those in the individual data. Instead it would be good to account for the heterogeneity in the background hazard. To implement this:

• backhaz would have extra columns, e.g. agegroup, sex, corresponding to columns of the individual-level data.

• agegroup would represent age at time zero. For each agegroup , backhaz would have multiple rows, giving the background hazard at different values of time since time zero.

• Each individual in data must have a background hazard value in backhaz that matches their covariates.

• Possible syntax survextrap(..., backhaz_strata=c("agegroup","sex"),...). backhaz_strata would default to NULL, meaning no stratification of the background hazard.

[chjackson]

Comment if you would make use of this!

[chjackson]

Implemented in https://github.com/chjackson/survextrap/commit/ba4330cec2ad0e5c5e0bbd07a62adf16b49ac39b