Implement metrics of algorithm/model performance

[ProbablyFaiz]

We're getting to the point where it would be good to have empirical metrics of our algorithms' performance relative to each other and third party techniques.

These include things like link prediction and other common statistical evaluation techniques like recall.

Some examples:

• https://arxiv.org/pdf/2106.10776.pdf
• https://github.com/VHRanger/nodevectors/blob/master/examples/link%20prediction.ipynb