TIP4P oxygen frame: primitive rings and CHILL+
input/traj/exampleTraj.lammpstrj, atom type 2, 250 oxygens
box 40 x 40 x 180.20265674635874 angstrom, orthorhombic
source_tree: 53fd1909724f9246db2d9fab0c8bfa52996f3a13
source_tree is git rev-parse HEAD^{tree} of the commit that contains
the sources compared below, recorded before this log was added.

Neighbour graphs
  cutoff: oxygen pairs with minimum-image distance <= 3.5 angstrom
  hbond: pydseamslib 2.11.1 geometric hydrogen bonds on that cutoff
         list (acceptor-H <= 2.42 angstrom, O-O-H angle <= 30 degrees,
         hydrogen LAMMPS type 1)
  The cutoff list has 509 undirected edges, degree 2 to 6, mean 4.072.
  The hydrogen-bond list has 460 edges, degree 2 to 4, mean 3.68.
  Every hydrogen bond is also a cutoff edge. The two lists differ by
  49 cutoff edges that are not hydrogen bonds.

Independent primitive-ring count
  A cycle is primitive when, for every pair of its vertices, the graph
  distance equals the shorter arc (Franzblau, shortest-path criterion).
  Simple cycles of length 3 to 6 were enumerated with the lowest index
  first and one direction, then tested against all-pairs distances
  capped at the cycle length. No engine ring routine was called.

  graph     edges   size 3   size 4   size 5   size 6
  hbond       460        1      230        4      192
  cutoff      509       37      273       34      167

  pydseamslib 2.11.1 ring_census on the default read (bonded auto, which
  is hbond for a type-2 dump) is 1, 230, 4, 192.
  The same package with bonded="cutoff" is 37, 273, 34, 167.
  The in-tree hydrogen-bond census matches the reference and 2.11.1.
  The in-tree cutoff census matches the reference and 2.11.1 bonded=cutoff.

  The 37 triangles are primitive rings of the cutoff graph, not a
  failure of the shortest-path test. One of them is oxygen indices
  0, 51, 54 with minimum-image edges 2.763, 2.860 and 3.458 angstrom,
  all inside 3.5. They are absent from the hydrogen-bond list, which
  is why that census has a single triangle.

CHILL+ on the same cutoff neighbour list
  pydseamslib 2.11.1 chill_plus uses neighListO and does not require
  the cutoff shell to hold exactly four neighbours. It reports 12
  interClathrate and 238 water. Three of those interClathrate labels,
  oxygen indices 28, 221 and 226, each have five oxygens inside 3.5
  angstrom.
  The published CHILL+ table requires a crystal or clathrate row to
  have exactly four neighbours in the cutoff shell. The in-tree
  cutoff path applies that test, so those three sites are water:
  9 interClathrate and 241 water. Both totals are 250.
  The tutorial expectation is the published count.

sI guest cages
  input/traj/genice_sI.lammpstrj, atom type 1, cutoff signature 512.
  Two cages, vertex lists of length 20 and 20, 40 distinct indices.
