Cluster analysis + NetworkX

[RudolfWeeber]

• Cluster analysis should make use of the short range loop for neighbor finding, if the cutoff is short enough
  • add a cutoff() function to the pair criteria, i.e, the maximum distance of a potentail neighbor
  • When running the cluster analysis for all pairs, check if the cutoff is smaller than the cell system range
  • If yes, use the neighor finding rountines from cells.cpp to iterate over all pairs of particles within the cutoff
  • otherwise, use the present code, retrieving all particles to the head node
• Cluster analysis should collect/keep edge information, if requested, e.g. in a std::vector<std;:pair<size_t,size_t>> storing paris of particle ids for neighboring particles
• Cluster analysis should be able to link/export to the networkx Python package (using particle ids as nodes and the edge info from the previous step)
• The per cluster "observables" such as longest distance and raidus of gyraiton should be made Lees-Edwards safe

[RudolfWeeber]

Step 1: Add a distance cutoff to pair criteria

• the pair criteria are in src/core/pair_criteria
• Add a virtual abstract method double cut_off() to the PairCriteroin base class
• Implement the method in the derived classes:
  • for DistanceCriterion the cutoff equals the distance
  • the eneryg criteiron has a cutoff already
  • one needs to be added to the bond criterion. Default shoudl be -1, but should be overridable
• For the bond criterion, the cutoff should be accessible via the script interface from Python. See src/script_interface/pair_criteria/EnerygCriterion.hpp for a template.