Explicitly provide negative facts

In the current form, we can only provide true facts and Ampligraph generates negatives automatically. In many cases, though it is very common to be certain about facts that are false. It would be nice to be able to inform the model. In fact it might be easier if when providing ground truth to set a belief score next to the triplet. For example

`nick friend Judy 0.7`


