kurtosis and skewness from a Histogram

[Peter]

The functions skewness and kurtosis take Vector which is to convenient when having a histogram with some bins with millions of data points. For mean and variance we nicely provide the Averager class that can calculate these moments quickly using the second argument in add(double x, int n). I suggest something similar for third and fourth moment.

[Peter]

I plan to create a class Averager4, similar to Averager but with more info (third and fourth moments)

[Peter]

(In [2799]) first version of class Averager4. refs #705

[Peter]

(In [2803]) fixed spurious test failures as well as a bug in operator+=. refs #705

[Peter]

(In [2809]) closes #705. Create a common interface class for average classes (averager_base) that is extended by additional base classes, averager_base2, averager_base3, and averager_base4. Averager is an implementation of averager_base2, and Averager4 is an implementation of averager_base4. For completeness added also classes Averager1 and Averager3 that implement averager_base and averager_base3, respectively. Using curiously recurrent template pattern to avoid vtables which has the downside that runtime polymorphism is not allowed (but also not really wanted here).

[Peter]

(In [2810]) add missing files in r2809. refs #705