update model data collector to include counts of agents for each risk attitude

[rlskoeser]

If we update model data with total number of agents in each risk attitude, then we can use that to compare proportions relative to total agents (based on grid size), and have better numeric values for use use in regression analysis (use the same buckets as our categories for risk moderate, risk averse, risk inclined). This means we can skip collecting agent-level data for the large scale batch runs - so maybe add a parameter to the hawk/dove multi batch run script to control whether agent data is collected.

[rlskoeser]

implemented in #66