E.step: E-step of the EM algorithm
In pzwiernik/LatentClass: Estimation of the latent class model for discrete data

This internal function performs an E-step of the EM-algorithm

1	E.step(phat, theta)

`phat`	The array of normalized counts (sample proportions) of format (r1,...,rm)
`theta`	The parameter vector in form of a list of size m+1. The last element is the marginal distribution of the latent variable. The other elements of the list correspond to conditional probabilities of the observed variables given the latent one.