print sample clustering by default

[borauyar]

Whether or not the training was done in supervised/unsupervised mode, the samples can be clustered based on embeddings. The optimal clustering results should be printed to folder by the end of the training (for both training/testing samples)