Constructor for an interpolation which is the derivative of another Interpolation

[danmackinlay]

Currently we have hessian and gradient to evaluate derivatives of Interpolation objects. However, sometimes we wish to simply construct a whole new interpolation which is the derivative of another and use this. (also, more generally, we might want to construct interpolations which are integrals of other interpolations, but there are a couple of other details there).

It looks like this should be mathematically not too complex, but I got lost trying to understand how this would get the correct order and boundary conditions.

[timholy]

Yes, that's a reasonable thing to want. Again, a PR would be welcome.