More General Options for Objectives / Constraints in Container Objects

[electronsandstuff]

This PR allows the container objects to store "non-canonical" optimization data. That is, for problems that are not minimization problems with constraints such that g >= 0 is feasible.

• New members in Population objects:
  • obj_directions: string where each character represents maximization (+) or minimization (-) for each objective
  • constraint_directions: string where each character represents a constraint satisfied when g is less than the target (<) or greater than the target (>)
  • constraint_targets: numpy array of the target values for constraints
• Population random generation includes option for randomizing constraint and objective settings
• Population random generator is updated so that constraints lie in (-0.5, 0.5) instead of (0, 1) which was always feasible.
• Plotting, file saving reflects new settings
• The container notebook has been updated with examples on usage of new settings. Population examples were also changed to avoid from_random which doesn't provide the best example for new users to library.
• New tests for saving and loading previous file format versions. New notebook to save file version checkpoints.
• New test for feasibility on example individuals