Improve runtime

[hkariti]

• Accept a precomputed dist matrix instead of calculating ourselves
• Use numpy operations for faster runtime

Tested on 1000 samples of scipy's half-moons with noise=0.24 and seed=1. Benchmark results on my machine:

master branch: time=128.46s memory=172864 KB this branch: time=12.08s memory=164668 KB

[hkariti]

Haven't noticed #17 before. I didn't compare their changes to mine