Gamma distributions

[peterstangl]

The current implementation of probability distributions related to the Gamma distribution might be confusing for the following reasons:

• GeneralGammaDistributionPositive is not a generalization of GammaDistributionPositive since instead of the arguments a, loc, and scale, it has arguments scale_factor, counts_total, counts_signal.
• No equivalent of GeneralGammaDistributionPositive with vanishing background uncertainty exists with arguments scale_factor, counts_total, counts_signal.
• The argument background_variance used in GeneralGammaDistributionPositive and GeneralGammaUpperLimit is not the variance but the standard deviation.

[peterstangl]

All the above points have been addressed in PR #210.