The simulation calls act() on every Agent that it can
find in the grid (model). So an Agent will cease to act
if it is replaced in the grid, by either another Agent (that's what
happens when a Rabbit eats a Clover) or by nothing (null).
When an Agent is killed by another Agent, it simply won't be asked
to act again. But when an Agent dies during its act(),
you need to take care that it does not continue to act after its death
(as a ghost or zombie). So it probably makes sense to have a dying
Agent exit its act() method promptly (or perhaps after
carrying out its dying wishes).