[Feature] New Class Structure for MathAgent

Summary

Restructuring the MathAgent Class Structure

Explain your context

The current structure of mathematical optimization in eta_utility is really hard to understand. I want to propose a more OO way of representing the MathAgent's structure.

What is the current behavior?

An additional environment is being specified, said environment uses a different method signature.

What is the expected behavior? What is the goal of adding the feature?

Environments have a unique method signature, the Agent handles it's complexity of having a Model itself.