merging e-values

E-values are easier to combine than p-values, which is one of their selling points. For instance, by linearity of expectation, given e-variables $E_1,\dots ,E_n$, any convex combination ${\sum }_i{\lambda }_i{E}_i$ is an e-value, where ${\sum }_i{\lambda }_i=1$, and ${\lambda }_i\ge 0$ for all $i$. Note this holds for arbitrarily dependent e-variables. This is, in some technical sense, the optimal way to additively combine e-values.

todo