Now search_knn_dist returns the distances with the point itself

[harish1996]

Adds the distances also with the point itself during calls of search_knn_dist. Calculating distances of each of the nodes again is redundant. If one doesn't need the distances, they can just ignore it and just form a new list using list comprehension.

only_nodes = [ i[0] for i in results ]

[coveralls]

Coverage decreased (-0.05%) to 86.268% when pulling 63c586d5d3d462db7a1915919d9fd2b6de0482ed on harish1996:master into 587edc7056d7735177ad56a84ad5abccdea91693 on stefankoegl:master.