Return data from od fit methods

[mauicv]

• Currently when you can fit on gmm and svm detectors the user can specify verbosity, however, we don't return any logs of the fit process despite computing them internally, These should be returned.
• Also, make sure that returned data is consistent.

[jklaise]

Do you mean actually return some object with the logs or just fiddling with the logging levels?

Bigger than this issue, but it's a great point to consider how we could expose application level logging. From our point of view, we just need to sprinkle logging statements, but then it's up to the user/application to define where to log and which level to expose (as in, our verbose) flags would not affect these. Good to check with scikit-learn as they have verbose flags too.

[mauicv]

I mean some object with the logs although perhaps logging would be better. Specifically in the case of gmm we allow a convergence threshold to be set. The training ends on the basis of this criterion being met or the training process running for max_epochs. I felt this should be returned to the user so they can check if this has occured or at least set this value on the detector somewhere.

[jklaise]

OK I see, it makes sense to both return something that can be programatically checked and also logging progress, both of which are kind of orthogonal.