[tests] design better tests on heuristics of bimodal distros or cluster to target

[gph82]

Many tests are failing because one method is supposed to find the two centers of a sampled bimodal distribution. The criterion of what constitutes having found the center should be relaxed or a more deterministc test should be devised

[gph82]

adding "cluster to target" to the list of improvable tests