Resolve expected vs. probability for mutation ops

[markcoletti]

Mutation operators take on an expected parameter that dictates the "expected" number of mutations per individual. Jeff pointed out that there may be scenarios where someone would prefer to specify the probability, instead. However, clearly "expected" and "probability" are mutually exclusive, and so steps would have to be taken in the design of these mutation operators to ensure that users don't accidentally misuse them by specifying both parameters.

[markcoletti]

I had posed some possibilities:

• raise an exception if the user tried to use both
• have a wrapper function where the user would specify the "expected" number of mutations that would calculate the actual probability, and then call the actual mutation operator with that probability
• have a class implementation of mutation operators that used either subclassing or overloaded methods to do The Right Thing

The simplest would be the first, but that feels kind of like a kludge. The second is more elegant, but adds some complexity. The third was added more for completeness as I think it's the worst of the three. :(

[SigmaX]

ex. Gaussian mutation doesn't have a p_mutate parameter.