Optimization ideas for mxpedigree

[smasongarrison]

• Consolidate/reuse relatedness matrices when they are the same for multiple families. right now each group gets its own copy which is hugely expensive.
  • you can reuse matrices if they have the same IDs, so we would need to overhaul a couple of things in the code so that the ids can be made more generic and then remapped.
  • See https://github.com/R-Computing-Lab/BGmisc/blob/temporal/data-raw/test_static_family_grouping.R
• reduce how much information is storred for the temporal matrices. Do we really need each of the following?

Object	Retain?	Reason
Eta_k	Usually no	Reconstructible linear predictor
L_k	Yes	Small and directly interpretable
K_k	Usually no	Reconstructible from L_k
V_k	Yes	Actual component covariance
V	Yes	Required by expectation