elliptical cylinder for axis ratio < 1

[pkienzle]

Elliptical cylinder has axis ratio in [1, inf] but the computation would support [0, inf], with radius minor not necessarily being the smaller radius. This would allow fits where one radius is fixed but the other could be larger or smaller than the first, but it also allows two solutions in the case where both radius_minor and axis_ratio are fit. If this is done, also change the random model generation to allow ratios from 1/max ratio to max ratio instead of 1 to max ratio.

Cylinder length is limited to 1 or greater. Look for equations that allow length to go to zero.