Reordered A/B in minimizeRmsd() to be consistent with supplied arguments; does not affect return values

a) some intrinsic flexibility associated with the first crossover
b) the spring for the twist of the helix
c) intrinsic flexibility associated with the second crossover

1) Proper accounting for a single bond representing a double crossover
2) Parallel-helix dihedral angle spring constant is now computed from three "springs" placed in series that represent:
   a) the flexibility of the bond on one side of the crossover
   b) the intrinsic flexibility of the crossover
   c) the flexibility of the bond on the other side of the crossover

Fixed bug introduced in commit 47b578f2 where ArbdModel.__init__ arguments were incorrectly provided by SegmentModel.__init__

Added test to check that model_from_basepair_stack_3prime works with circular ssDNA; removed warnings and made function work with lists instead of numpy arrays

Fixed polygon_mesh reader so that it works with user-generated .ma files, not just autogenerated ones

find_stacks routine moved to segmentmodel_from_lists using only coordinates and transforms; generate stacks automatically if None

kept old find_stacks routine in segmentmodel_from_pdb as find_stacks_mdanalysis

1) Passthrough of mrdna.segmentmodel arguments
2) Improved input validation
3) Stacks are removed unless stack[bp[stack]] == stack
4) Cleaner code for assigning ends of terminally-connected "segments"

Eliminated numpy log(0) error message when writing bond potentials, which is caught anyway by .isnan

Prepared code structure in potential section so that dsDNA and ssDNA are treated using separate defs