Genetic value API

[molpopgen]

Recent discussions elsewhere demonstrate that the specification of the genetic value objects is confusing. Specifically, the ndemes field is a kludge.

Ideally, gvalue would be a list of one object per deme. The dimensionality of each object has to line up with the dimensionality of the effect size distribution. In other words, multi-deme models with pleiotropy** and multi-deme models with correlated effect sizes across demes need to be conceptually unified.

**do we even support this?

[molpopgen]

"Warning": any change to this requires mucking around on the C++ side.

[molpopgen]

To avoid breaking API we'd have to deprecate the current ndemes argument.