Eval l5pc test unstable

[wvangeit]

This commit: https://github.com/BlueBrain/BluePyOpt/commit/71a94b0ffa853133b688bcf4ef1755c9de415426 disables the comparison of the values in the dict returned by the l5pc evaluator. This was also the case when using nosetests. It seems these values are pretty unstable. When running on BB5 (trying different python versions) one always gets the same result. But when running on github actions the scores are different (up to 1 decimal change). This happens even when using compiled versions of neuron (exactly the same version).