Reduce dependencies

[famuvie]

We currently have the dependency structure shown below.

Most dependencies are actually via ggplot2. Should we produce the default plots using base graphics only? Maybe we could simply suggest ggplot2.

Another source of heavy dependencies is pedigreemm, which in particular requires nloptr which in turn requires an external library that needs to be compiled and is often problematic. This looks too problematic, given that we only leverage the pedigree class in pedigreemm and the functions getA() and getAinv().

[famuvie]

For getting rid of pedigreemm we should:

1. Use another class for representing pedigrees
2. Use alternative functions for computing the relationship matrix A and its inverse

For 1, we could get away with a simple data.frame (as other packages like pedigree or nadiv do) or create a custom class. Then we should adapt the usage of the pedigree throughout the code.

For 2, there are several options:

• the package pedigree provides fast compiled functions makeA() and makeAinv(). The inconvenient part is that it works by writing the results to text files which need to be read into memory.
• the package nadiv also provides fast compiled functions with the same names as pedigree and it does not bring in too many sub-dependencies. Moreover, it contains additional functions for computing dominance matrices, inbreeding and other potentially useful stuff.
• alternatively, we could just write custom functions for breedR, or borrow the code from the previous packages (also possibly from AnimalINLA) if licenses allow that.