SHARKFIn agent class extended HARK class with SHARK features

[sbenthall]

There are several features needed by SHARKFin agents that go beyond HARK agents.
These include:

• A store of number of shares owned
• Some metadata that helps with the fact that we are using daily market returns while HARK agents generally operate quarterly or annually -- this is handled by the macro_day property

Currently, these are handled through monkey-patching / direct variable assignment on the AgentType objects by the simulation code.
This is a mess because these interventions are ad hoc and there's nothing except the simulation code logic to maintain the necessary invariants in the underlying data structures.

We should have better containment on this functionality, and locate it on the agent objects themselves, so the simulation code can focus on the simulation loop structure not the agent manipulations.

[sbenthall]

This SHARKFin agent should also be what knows how to use the new solution object.
This can go in sharkfin/population.py

[sbenthall]

Methods to move from sharkfin.simulate to sharkfin.population (?) or otherwise a wrapper/extension around the AgentType that constitutes the AgentPopulation:

• attend() -- updates the risky expectations from the financial model; computes new share demand, returns the difference between currently owned shares and demand.
  • https://github.com/sbenthall/SHARKFin/blob/master/sharkfin/simulation.py#L141
• compute_share_demand() -- uses the current risky expectations, accesses the master solution (?) and gets the number of shares needed.
  • https://github.com/sbenthall/SHARKFin/blob/master/sharkfin/simulation.py#L164
• macro_update() -- for a particular agent, run their earning and consumption state transitions, but NOT the risky asset transition. If they had to sell shares, execute that trade
  • https://github.com/sbenthall/SHARKFin/blob/master/sharkfin/simulation.py#L238
• update_agent_wealth_capital_gains() -- given new price, dividends, ror, update the asset levels of the agents.
• https://github.com/sbenthall/SHARKFin/blob/master/sharkfin/simulation.py#L444

[sbenthall]

@alanlujan91 to move the methods/functionality over in a nicely (probably re-) designed way into the AgentPopulation or SHARKAgentType or whatever class.

@sbenthall to wire this into the AttentionSimulation afterwards.

[sbenthall]

To make this easier (hopefully), I'm going to work on doing this refactor on the the existing/old AgentPopulation class, as this overlaps a bit with #88. I'll be adding these changes to PR #113 for your review @alanlujan91