Add the quasibinomial model for pairwise comparisons

[pawel-czyz]

We have quasimultinomial model, which tries to fit all the growth advantages at once. There exists a simpler model, which targets only the proportion $p(t) = \frac{y_1(t)}{y_1(t) + y_0(t)}$ and estimates a growth advantage directly, using a two parameter model $p(t) = \mathrm{logit}^{-1}(g t - b)$.

Prepare the functions that:

• Allow one to calculate $p(t)$ and plot it.
• Fit the quasibinomial models using maximum quasilikelihood and quasiposterior.
• Function for calculating overdispersion.
• Assessing the residuals.