Add "reservoir host" experiment

This adds a new simulation experiment intended to investigate the "reservoir host" hypothesis, i.e. look for patterns in the prevalence values when some species are competent hosts that are susceptible and infectious whereas others are only susceptible and will not measurably contaminate visited flower patches. At the moment, this is modelled by setting a contamination rate of zero for one out of two otherwise identical species.

This also requires that infection/recovery/contamination/decontamination rates become specific instead of global parameters. Therefore the memoisation strategy is adjusted to keep all per-tick probabilities (which are costly to compute due to the exponentials involved) in a global hash table instead of storing them within each flower patch entity.