Improve spline resampling to use any degree spline fit

We now have a spline resample implemented in specutils, but it's only a cubicspline. There is a good scipy function for doing a spline fit of 5 or less degrees (`scipy.interpolate.UnivariateSpline`), but, it cannot handle the "edges" of the fit by filling with NaNs without some extra helper code. `CubicSpline` however, has built in NaN filling. For time reasons, since the current implemented default fill is NaN, we're using `CubicSpline`.

Once #489 is done, we should change this out accordingly.

Read the existing spline resampling implementation in specutils and review #489 first. Replace the cubic-only fit with scipy.interpolate.UnivariateSpline for degrees up to five while preserving NaN filling at the fit edges; done means the resampling supports those degrees and retains the current default behavior.

Written by the indexing model from the issue text.