Population date of birth initialization should take place upon initialization of birth or mortality functionality, not overall model initialization.

[KevinMcCarthyAtIDM]

Right now, the code flow has date of birth initialization taking place during the model initialization.
At this point in the code flow, the model doesn't know which demographic steps are included in the model components, and the initial date of birth distribution does depend on those. So I'd suggest that we move this over into the initialization of the demographic steps.

The right place to do this would be during initialization of one of the demographics steps. I'm doing constant population demographics, with birth rate = mortality rate, so it makes the most sense for me to not have a mortality step and initialize the "dob" distribution of existing agents when I initialize the births model component.
In a fullest implementation, the initial date of birth distribution is dependent on both the historical birth rates and the historical age-dependent mortality, so needs to know the parameters and dynamics associated with both. In specific implementations where one or both of those might be constant over time, age independent, etc., it could live either on the births or mortality initialization, whichever makes more sense.