I want to use a custom representation that consists of a GP tree,  
plus an additional real number (both the GP tree and the number are  
used in the fitness evaluation).  Besides implementing a custom  
individual which extends GP individual,  what else should I  
implement?  I want to perform crossover and mutation on the tree as  
per standard GP, and gaussian mutation on the constant.  Do I need a  
custom Mutation and Crossover pipeline?  What about selection?