hawk/dove myopia: look at neighbors behavior over more than one round

[ADDED 11/8]

A different way for an agent to determine the probability that some randomly-selected agent plays HAWK might be to look at what her 8 neighbors have done on every round so far, or on the last 10 rounds, or whatever. This would be fairly easy to implement, I think. An agent with, e.g., r = 5 would require [total HAWK among neighbors in past y rounds] > 5y
