Update predator preference during the simulation.

[ismael-lajaaiti]

Ecological context

By default, we assume that a predator split their time equally between the number of its preys. Therefore, the predator preference for each of its prey is $\omega = \frac{1}{\mathrm{p}}$ where $p$ is the number of prey of the predator. The thing is that $p$ is not a constant, because species can go extinct $p$ is a function of time. To be more precise, we should then write $p(t)$. And the way we define predator preference currently is $\omega = \frac{1}{\mathrm{p(0)}}$, that is we use the number of prey at $t=0$ to define the preference. However, there is a number of argument to say that the predator preference should change when one of its prey goes extinct, therefore the preference would become $\omega(t) = \frac{1}{\mathrm{p(t)}}$.

Implementation

I see two ways of implementing this feature:

1. Through a callback triggered on species extinction that would update the preference matrix.
2. Re-define a preference matrix not as constant, but as a function of species biomasses. Thus, we remove the contribution of extinct species ($B=0$) to the preference.

[iago-lito]

Interesting. I can also think of the following one 0:)

3. Suggestion: remove extinct species from the simulation and restart solve() on every extinction. #80

If not the chosen solution, then having parameters values depend on the current number of live species is something to be considered when investigating #141. An ideal solution to #141 should make it easy to feature what you request here :)