Added the one-dimensional background estimator from the ROOT TSpectrum

[cl-fontana]

Here is the "distilled" function to estimate the background from a one-dimensional spectrum.
It was extracted from the TSpectrum [https://root.cern.ch/doc/master/classTSpectrum.html] source from ROOT [https://root.cern.ch/].
It now depends only on standard C++ libraries.

[wojdyr]

Thanks!
Now I'm thinking about the command line interface. For Shirley background we have:

@0 = @0 - shirley_bg(@0)

I'd do something similar for this background. Can you suggest a name for it?
I suppose there is no way to get in touch with Miroslav Morhac and ask him for opinion?

And how to handle all these parameters?

                               int numberIterations,
                               int direction,
                               int filterOrder,
                               bool smoothing,
                               int smoothWindow,
                               bool compton

Which of them should be exposed to the user? What are sensible default values?

[cl-fontana]

We have been using that algorithm for spectra that look like the attached picture (which was fitted with the manual background). For this experiment we have rather broad peaks with a FWHM/centroid from 5% to 12%.
For that data I have been using this set of values:

1. number of iterations: ~15
2. direction: decreasing window
3. filter order: 2
4. no smoothing
5. no Compton estimation

You could try to use that set of values for some data of yours and see if you get reasonable results. If so we can call this our set of defaults.