class calculating likelihood ratio test for binomial models

[Peter]

The calculation is straightforward for the individual ML and the joint one with condition p1==p2. Possibly we should add a p-value using Wilk's theorem (for large samples).

Possibly one could also add a weighted version, which in this context means a "fuzzy binomial", i.e., one is not sure want the coin toss showed but believes it be be tail with 90% certainty. Not sure how the math works out in that case though, so will start with unweighted case. Just mention the weighted case so the interface is designed with that in mind.

[Peter]

(In [3134]) refs #769 skeleton for new class LoglikelihoodRatioTestBinomial? (not yet functional).

[Peter]

(In [3138]) change name of class and implement function llr(). closes #769