Feature/hmm export testset

[ChataingT]

In regard to the ISSUE #14.
Here is a proposition to solve it.
The solution take part in the "betman segment_motifs" feature :

• A boolean to save the hmm model
• An input path for a testing dataset that will be used only in hmm.predict()

[gchindemi]

Thank you for the contribution! I think the proposed solution can be further improved.

First, there is a minor bug: test_predictions is computed but not returned.

Second, the logic could be adapted to better support labeling the test set using a pre-trained model, as proposed in #14. To achieve this, I suggest modifying the function signature to allow the user to specify the path to a pre-trained model, and drop the test set component. That is, if the user wants to label another dataset (e.g., a test set) using a previously fitted model, they can simply specify the path of such model and call the segment_hmm function as usual.

Additionally, it would be important to clearly warn users that any HMM configuration parameters (e.g., num_states) will be ignored when loading a pre-trained model. A log message should be more than enough.