[metrics-team] New tool to rank relay similarity

[Philipp Winter]

To be able to leverage Stem's API, I reimplemented sybilhunter's nearest
neighbour ranking in Python.  The tool takes as input a relay
fingerprint, and then ranks all other ~7,000 relays by their similarity
to the reference relay, most similar one first.  Relays are similar if
they share information in their descriptor like ports, or their
bandwidth values.

The code currently resides on GitHub, and soon also in a Tor user
repository: <https://github.com/NullHypothesis/trnnr>
There's also a README file that talks about usage examples.