add support to fetch trained logistic regression model in the logreg metric

[MayankAgarwal]

fixes #58

Adds a property to get trained Logistic regression models for a fitted Logistic Regression Sensitive Subspace metric.

metric = inFairness.distances.LogisticRegSensitiveSubspace()
models = metric.logistic_regression_models

The models variable will now contain a list of logistic regression models modeling P(K|X)

Note: cannot add this to the state of the metric since state_dict only supports tensors to be specified, whereas this is a sklearn object.