Stats for K, S and pairs

[patricoferris]

Where K is the sampled pixels in the project, S is the candidate matching set and pairs is the X number of matching pairs (currently 100, but see #41).

Talking with Tom, it would be very useful to dump some stats for K, S and the pairs. In particular, the standardised mean difference which allows us to compare say the mean slope from the pixels from K and the mean in the pixels from S and how similar they are in units of standard deviation (apparently 0.2 and below is better in general).

For this I think we want to maybe adopt the quick DUMPDIR approach I started in ba2d0ef and add this to the output of say find_pairs. If they are CSVs I can render them directly in the browser too.