Add test for not round percentage selection

[guillep]

          Actually, my point is also that if we change the number of mutations the test will break.

Take my example of 33 mutations, and let's suppose we chose 20% rounded up, running 7 mutations (6 or 7 here is annecdotic).

Then the assertion

analysis generalResult numberOfEvaluatedMutants / numberOfMutations * 100

Will be equals to 7/33*100 == 21.21 which is not your expected 20 :)

Equally, if we choose to run 6 instead of 7, we would have 6/33*100 == 18.18.

At the end the question is: do we stop before exceeding the budget or after?
So we need a test for this ^^

Originally posted by @guillep in #34 (comment)