- started with split-method for SSV objects
- fixed circular import

TODO:
- add all functionality necessary to split/merge SSVs
- add a mechanism to auto-detect which classification should be rerun in case of SSV modifications (e.g. add model paths to config -> infer all types for which a model is available)

- refactoring

Merge branch 'dense_segmentation' of github.com:StructuralNeurobiologyLab/SyConn into dense_segmentation

Merge branch 'dense_segmentation' of github.com:StructuralNeurobiologyLab/SyConn into dense_segmentation

Speed-up rep coord calculation dramatically. Added numpy temporary random seed handler to basics module.

# Conflicts:
#	syconn/handler/logger.py

- bug fixes

# Conflicts:
#	syconn/handler/prediction.py

- adapted histogram normalization in 'extract_skel_features'
- minor adjustments

added object extraction warning if specified overlap is bigger than cset overlap which leads to stitching failures