Implement IBD algorithm on Spark

[jtarraga]

Identity-by-descent (IBD) is a good estimation to find pairs of individuals who look too similar to each other. IBD scores are computed from the identity-by-state (IBS) scores. Note that the IBS algorithm is already implemented in the oskar package.

IBD will be implemented as a Spark dataset transformer.

IBD implementation will be based-on PLINK, http://zzz.bwh.harvard.edu/plink/ibdibs.shtml#genome
PLINK package will be used to validate our results.