Package motivation

[SouthEndMusic]

Refactor: remove TaylorDiff as a dependency, and let the input of the evaluation and rule computation be a function which gives the first 2s derivatives of the integrand at the input x (either in place or out of place?).

Motivation: This package is useful for numerically integrating functions for which computing higher order derivatives can be cheaply implemented. This then has a preference over a gaussian quadrature rule with more nodes. Use the sine function as an example, first evaluating the quadrature rule with TaylorDiff and then with hardcoded derivatives.

[SouthEndMusic]

The refactor was done in #14. Storing the derivatives in a preallocated cache: #15.
This issue is now only for the motivation part.