Added computing of R4 (sensu Suzuki et al 2005).

This consists in dividing counts by the maximum across synonymous codons
for a given amino-acid, and is supposed to avoid certain biases in PCA.