Merge pull request #2595 from RosettaCommons/aleaverfay/pdb_conect_performance_bugfix
JD3 bottleneck / PDB output taking too long
Attempting to fix a pretty serious performance bug in PDB writing that spent O(N^2) time trying to determine the CONECT records. This impacts JD3 protocols quite heavily in that it creates a bottleneck on node 0 when outputting the results from other nodes.
Fixing a few bugs in the way CONECT records were written out at the end of PDBs. For this reason, lots of integration tests change, but not coordinates of structures written out -- only the record of which atoms are bonded to which other atoms.