Performance minutiae

[James-E-A]

From the original paper:

The compactness of the recommended solutions is due to the use of nested equations to compute elliptic terms and of only three trigonometric function: sine, cosine, and arc tangent. Nesting reduces the number of operations involving storage and retrieval of intermediate results (particularly when programming) in assembly language), reduces the length of the program and the time of execution, and minimizes the possibility of underflow.

I'm marking this PR as a draft, pending benchmarking verification that these changes are actually relevant given the introspective ability of today's compilers, but I strongly suspect it's at least probable that it will make a difference and take full advantage of the design that these equations have had since the start.