Base distribution ablation MNIST

[fariedabuzaid]

We should define an Experiment that tests uniformly scaling flows with different base distribution regarding their fit for the AD task.

• Define HyperOptExperiment and train models
• Comparison:
  • AD performance ROC AU
  • QQ-plots

[giorgiapitteri]

Issue related to the PR: #5

Maybe we can check if the eval module is enough for this issue and merge it for the moment? Since this PR is getting bigger with different things inside.