v2.1.0: selectable precision

This release adds a selectable-precision evaluator, splinetable::evaluator_type<T>, where T is the arithmetic type used to evaluate the spline. This can be used with double to reduce round-off error w.r.t. the default single-precision implementation, which can be significant when evaluating derivatives of nearly-constant surfaces.

Full Changelog: v2.0.7...v2.1.0