fix bug with calculating helix rolls to minimize crossover strain

[dave-doty]

Don't have time to describe in detail, but set up a simple system with two helices and the "unstrain crossovers" feature doesn't seem to get the angle correct. Not sure if it's also happening in the Python package. For reference, this design has the problem:

[dave-doty]

I've got a bit more of a clue what's going on. Things look fine if there are no deletions/insertions. But there are deletions above, leading to the strange results. What I'm not sure about is whether

1. The deletions are not being properly accounted for, or
2. They are, and I'm just not thinking about how the assigned Helix roll angles are actually good, given the deletions.

[dave-doty]

Here's the design above (mostly) with and without deletions: roll-issue-with-deletions.zip