Improve DOFs and Objectives for better control of Pareto optimization

[thomaswmorris]

This PR does two main things:

1. We interpret objectives as either a fitness (to be minimized or maximized) or a constraint (to satisfy some condition). Optimality is then the optimal/Pareto-efficient fitness within all the constriants.
2. We generalize the notion of a transform, so that we can transform the inputs and outputs with any function we want to better describe them using a GP.

Some small things:

• A lot of processes are more robust and numerically stable.
• We delete a lot of spaghetti code.
• Add more to the documentation.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[mrakitin]

Please rebase on main to eliminate the commits from already merged #64.

[mrakitin]

Looks good overall, but it needs a few minor touch-ups suggested below.