MATH-1359: Create expm1(x)/x implementation

[RayDeCampo]

I know things are in flux and you may not be able to use this, but here it is anyway.

I'm not completely happy with the accuracy of the evaluations of the derivatives but I wasn't sure if it was worth trying to improve them.

Note I based this off of the develop branch at commit 827d964, right before the dependency on commons-rng was introduced because it was easy to get that to compile. I don't think it will matter since the PR only contains new files.