Add optional flag to the evaluator to save the estimated values in the tree scope for symbolic regression problems

[bburlacu]

Currently the estimated values returned by the tree interpreter are used (independently) by:

• the problem evaluator
• the phenotypic diversity analyzer/phenotypic similarity calculator
• the schema evaluator operator (part of the schema diversification algorithm)

Therefore, a lot of computational effort is wasted in recomputing these values when they are needed. For this reason it would make sense to add an optional flag to the evaluator (with the default value set to false) to specify whether or not the tree estimated values should be saved inside the scope for later reuse. This requires minimal changes to existing code and has no negative impact on performance in the default case. On the other hand it would make a huge difference for runs where phenotypic similarities are needed for analysis or algorithm tuning.

[bburlacu]

r12973: Implemented the necessary changes in the evaluators, and removed obsolete code from the phenotypic diversity analyzer.

[bburlacu]

r12977: Reverted changes to the evaluators as they impact performance and memory usage.