functions should be agnostic to 'id' and `the_geom` where appropriate

[andy-esch]

The original k-means function (https://github.com/CartoDB/crankshaft/blob/b547d35882dafe19870bd0428384bd22c63f87fc/src/py/crankshaft/crankshaft/clustering/kmeans.py#L20-L21) does not take an optional database id -- it must be cartodb_id. This should be flexible so the extension can be used on databases outside of carto more easily.