Test123 outputs unnecessary file with unstable content

[mdmitry1]

This test does not need a model, as its spec has system section:

"system": {
      "y1":"0 if p1<=0 and p2>-1 else p1",
      "y2":"0 if p2<=0 and p1>-1 else p2"
  },

However, its output contains Test123_smlp_toy_basic_sampling_predictions_summary.csv file with content changing from one run to another.

[zurabksmlp]

This test among other things tests an initial implementation of model refinement. We have a system, sample using DOE, then build the model and check its quality. The full refinement flow will find a solution using model, check it against the system, and if the solution is also solution for the system (say counter example to some assertion) then no further refinement is need; else there will be a refinement step.

These lines in Test123_smlp_toy_basic.txt confirm that DOE was run:

smlp_logger - INFO - Pareto optimization: End
smlp_logger - WARNING - factor p2 contains duplicate elements in the DOE spec; they will be dropped
smlp_logger - INFO - DOE table with 100 entries has been generated
smlp_logger - INFO - Reporting prediction results: start

The DOE itself in SMLP is not deterministic across runs and this is what we see in tests 123 as well as in at least some of the DOE tests 34-45. One option here is to ignore DOE non-determinism related diffs during regression, until it becomes important enough for us to investigate and achieve determinism. From what I can recall, I most likely tried to use seeds but it did not work at that time for me. Now that we have this issue opened, we can investigate deeper (with due priorities).