constrain randomization to repications within user-defined blocks

[mwibral]

In order to have take into account certain statistical covariates that are homogenous over certain subsets of replications, but that do vary from subset to subset, it would be good to be able to restrict permutations of replications to randomization within blocks of replications.

Suggestion: add a field to settings that takes a list of the form [start block1, start block2, ...]. This setting would be passed to the corresponding permutation routines of the data class.