[pyclustering.container.kdtree][ccore.kdtree] Reduce amount of Euclidean distance calculation

[annoviko]

Introduction
Refer to #369 looks like amount of lack calculation during search can be significantly reduced. Investigation is required.

Resources

1. Presentation about KD-searching - http://andrewd.ces.clemson.edu/courses/cpsc805/references/nearest_search.pdf.
2. Book - M.Samet. The Design And Analysis Of Spatial Data Structures. 1994.

Description
Use DBSCAN algorithm to test results.

[annoviko]

Implementation of optimized, balanced KD-tree is required, current implementation is original KD-tree. Optimized KD-tree should be implemented for python and CCORE.

[annoviko]

Current implementation is optimized and suitable for dynamical insertion and removing nodes, for example, for CURE. But in case DBSCAN and OPTICS there is no need to add new or remove old one nodes, static KD-tree is more suitable because it is balanced and ensure fast search procedure. Therefore new KD-tree should be introduced kdtree_static that will be used by OPTICS and DBSCAN.