Improve performance of 4th and 6th-order symplectic integration with spin

[cemitch99]

The current implementation of spin in elements using symplectic integration uses a three-term splitting to define a second-order method, which is extended to higher order using the Yoshida method. This is effective, but not efficient--many map evaluations are required.

It would be worthwhile to spend time exploring alternative symplectic integration approaches to improve performance by reducing the number of map evaluations required.