Implement Spline fitting for continuum normalization

[tzdwi]

Opening this issue to remind myself to do it when I have some spare time.

Currently continuum normalization is via a standard or a least squares polynomial fit. If you don't have a standard star, the least squares polynomial can run into some issues with molecular bands, or particularly broad lines. These issues become evident when trying to stitch orders together. I'd like to implement something like IRAF's continuum task, and fit a spline with iterative outlier rejection.

Thoughts and input welcome.