Parameter distribution across the agent population

[sbenthall]

A major feature of this REMARK is the realistic population model.
This involves a lot of hand-coding of elementary data structures in a way that is perhaps hard to understand and adapt.

https://github.com/econ-ark/DistributionOfWealthMPC/blob/master/Code/SetupParamsCSTW.py

Refactoring this into more generalizable components would be a win.

An example of an alternative way to define populations is the AgentPopulation class that Alan developed for SHARKFin:

https://github.com/sbenthall/SHARKFin/blob/master/sharkfin/population.py

My recommendation is that we use refactoring of the DoW population code as a vehicle for improving HARK's support for defining populations. Some version of AgentPopulation could be merged into HARK, which would improve downstream projects.

It's not immediately obvious how the SHARKFin AgentPopulation could be integrated into DoW, however, as the former is mainly about defining a population whose parameters range over a given mathematical distribution. DoW may alternatively drive the development of improvements to the HARK API for lifecycle models.