Modular IPM/MM for multi-strata simulations

[JavadocMD]

Many simulations involve models which separate individuals into separate but interactive strata -- this could be by age group, risk classification, species, etc. We would like epymorph to be able to handle these simulations as smoothly as possible.

Based on previous explorations, we believe we can take an even-more-modular approach than originally thought. That is, more than having an IPM/MM/GEO as modules in epymorph, perhaps the strata themselves can have their own IPM/MM/GEO configuration, and then we can find ways to express the interactions between strata. This might solve many of the possibly thorny issues that were coming up in how to structure this using singular IPMs and MMs.

Example 1: if our model involves humans, birds, and mosquitoes, we can select an IPM for each species, then a movement model for each species. This is very natural, since each species has very different behaviors in both movement and disease mechanics. However, these IPMs do interact -- infected mosquitoes infect birds, and infected birds infect mosquitoes -- so we will need to be able to describe the linkages ("meta edges") between them.

Example 2: many diseases have disproportionate effects on people of different ages. So you might break up the population by age group and apply slightly different mechanics to each. Movement is the most obvious -- adults tend to commute to work, while children tend to attend schools -- but there may be IPM differences as well. In this example, the IPM interaction network is expected to be much more dense (every strata impacts every other strata), but it can still be expressed as a collection of meta edges.

(There are additional details to work out regarding the initialization routine, data binding, ADRIO support for common needs, and so on.)

This task is to explore an implementation of the above.

[JavadocMD]

This effort is being temporarily paused (back to backlog); we have a working preliminary implementation in the multi-strata-sim branch.

The reason for the pause is to slip in a few more features in the current 0.4 line (non-breaking) before we integrate this one in the 0.5 line (minor breaking changes). These features will be of immediate value to folks using 0.4 and will also help us test and polish the multi-strata features.

Namely, those features are: 1) attributes declaration in the movement model, and 2) optional detailed movement data collection. We want #1 for multi-strata because the parameter input is now more complicated and would be nice to have some tool support for it. We want #2 because movement is now more complicated and is difficult to verify without detailed information about the inner workings of the system.