Fix multiple issues related to strand generation, also reduce timestep to 100fs if coarse_local_twist==True

Added fill_sequence parameter to SegmentModel.set_sequence and to the mrdna and enrgmd scripts, with default value 'T', since that is most common for origami. Thanks to Elija Feigl for recommending the change.

Connections between ssDNA and dsDNA should now always be of the type 'sscrossover', never 'intrahelical'

The latter could result in bonds between the end of dsDNA and ssDNA
from forming if the dsDNA end was also bound to another helix, but
this may remain an issue for three-way dsDNA junctions.

Added box_type=orthogonal and CUDA_DEVICE={gpu} to oxDNA input file; thanks to Elija Feigl for bringing the issue to our attention

Ensure no duplicate bonds in ENM, logging violations; thanks to Elija Feigl for bringing duplicate bonds to our attention

Updated ENM to avoid situation where network is applied to single-stranded DNA at end of double-stranded helix

Fixed two small bugs, one involving use of is for string comparison, the other involving an incorrect function call when a restraint is added to the system with only one argument

Fixed model_from_lists for 3'-to-5' structures; fixed issue with single-nucleotide 'crossovers'; added debug flag for tracing strands

Made cadnano reader set azimuthal angle more accurately; some problems could arise with cadnano2 files that are saved in cadnano2.5 due to incorrect assignment of eulerZ in cadnano2.5

Removed os.sync() lines prior to running arbd/oxDNA simulations because they can hang indefinitely; needs testing