Optimal binning

Some people do not like that at high energies the spectrum
looks not as nice as with sn binning. Added a qualifier "min_counts"
to rebin to at least this counts (but used inside optimal binning
algorithm).

Be warned that this destroys the purpose of the algorithm (at least
in the affected bins) and does not optimize between minimal resolution
and maximum information.

Also, most of the time spectra look "bad" at high energies is because
the background is significant, producing negative numbers when subtracted.