Tests for ABCD experiment

Reproduce results in the report.

Reproducing exactly results from the report is not important. Try 3, 5, 8 equally-distributed time points instead of 10, 20, 30 - that was uninteresting (data already prepared). IMVHO 5 should be enough and similar to 8, while 3 might be insufficient.

Multiple-runs test for each data-sheet, with few auditing assertions like: projection 0000100 expected (in each), projection 0001111 not expected / expected (cf. Fig. 6, 9 in the report) etc. Then also bulk tests for each tested aspect: std dev, nr of params, nr of samples. These would have bulk assertions like: in total more projections found, or, more specific, projection 0000010 detected more w/ nr of samples, and these would be the one to also produce the visual inspection (*).

The issue will be a running time so one has to try to keep it reasonable, e.g. five runs per case, low samples size for std. dev. and nr o points, by default only two cases per group: 5% vs. 15%, 3 vs. 8 points, 64 vs. 128 samples etc.

(*) Beware: plotting in tests is tricky. I suggest to make it as a part of report, but then you need somehow pass results files paths to from the tests. If you decide to plot in test via diagnostics mechanism, you can assume for convenience compatibility w/ MATLAB R2014b or higher (otherwise don't plot).