Set up library configuration with setup.py and minor changes to source

[fpier21]

Brief Description

This pull request introduces the possibility of installing the repo as a library, and it updates the code for compatibility purposes. Main motivation is to allow the usability of the code also with updated dependecies.

• Configured the repository to be a library
• Updated the requirements versions and fixed compatibility
• Added the option of annotating only a percentage of points in the plot
• Added the auto-increasing of acceptance threshold (probably an old bug)

Change Description

Added a setup.py file and refactored the project structure to be installable with pip install .
Removed dependencies versions because of compatibility issues. Only nlopt must be pinned to be compatible with the code.
Added parameter "annotate" in plot method to choose the percentage of points to annotate (float from 0 to 1). Default changes to None (do not annotate).
If failing to find to an initial decision boundary (i.e. the model is confident about its predictions so that there are no points in a neighborood of 0.5) auto-increasing the acceptance threshold (range around 0.5 to consider the decision boundary points) with a step of 0.03. Fixed bug: keep trying to find the boundary in endless while loop.

Testing

Local Testing: I tested the library and the functionalities on my local linux machine and it worked as expected.