in dimension above 38 or so, gradient injection on sep-CMA triggers stagnation criteria

[beniz]

Here is how to reproduce:

• sep-aCMA-ES run in 100-D

./test_functions -fname fsphere -alg sepacmaes -ftarget 1e-8 -dim 10

• sep-aCMA-ES run in 100-D with gradient injection (problem is here):

./test_functions -fname fsphere -alg sepacmaes -ftarget 1e-8 -dim 100 -with_gradient

this yields something similar to:

INFO - stopping criteria stagnation => oldmedianfvalue=512.312040743035 / newmedianfvalue=583.127653408428

• fix by deactivating stagnation criteria:

./test_functions -fname fsphere -alg sepacmaes -ftarget 1e-8 -dim 100 -with_gradient -no_stagnation

The fix is obvious while the underlying reason of the problem is less clear. This probably needs to be investigated.

[beniz]

Related to #97, closing.