Improve RL training and implementation workflow.

[AgustinGrillo]

• Re-test Optuna.
• Improve agent saving workflow. Automate model saving process.
• Train new RL agent with new Balancio version (0.6). Try with different observation space.