[WIP] Lucas MockMarket and USUAL setup

[sbenthall]

Working on #182 ...

• this refactors how populations are configured and passed through to the run_any_simulation.py script

Still need to implement #184

And need to pass appropriate values into USUAL expectations.

[sbenthall]

Added a function for computing price/dividend ratio based on Eq. 21 of these notes, but with one hiccup: 3da0fa1

I think that the (daily) dividend standard deviation and the agent's (quarterly) discount factor are in different time scales. I'm not sure if that means the equation needs to be adjusted. Any thoughts @alanlujan91 ?

[sbenthall]

TODO for this PR:

• [x] Compute the RiskyAvg and RiskyStd appropriate for Lucas with given $\beta, \rho$ and pass through to both the AgentPopulation and the UsualExpectations module. This may mean accommodating the differences in daily/quarterly scaling.
• [x] Generalizing the build_population script to work if RiskyAvg is not an agent-varying parameter as in the LUCAS0 case as well as when RiskyAvg is age varying, as in the WHITESHARK case.

I think I need to at least talk to @alanlujan91 about this, whenever he becomes available, before pushing further ahead on this.

[sbenthall]

Tests fail because of upstream changes to HARK -- approx -> discretize

[sbenthall]

Recent commits have done the following:

• add an 'ex_post' parameter which lists explicitly the ex post heterogeneous parameters over which to merge the solution. For the LUCAS0 population, this is None.
• Fix the conditionals that check for out-of-bounds risky_share values in compute_share_demand() to work when AgentCount > 1 (it was assuming scalar/1-valued arrays before)

But there are still some issues that are not fixed by this.

• The AgentSolution code really is written assuming that there will be some ex post continuous parameters to merge over. It chokes when you try to pass it None or an empty list to merge_by. I've tried to fix this myself but have gotten to a point where I need to engage @alanlujan91 on this.
• This is working locally because I have an older version of HARK in the environment, but as you can see tests fail on this PR because of changes to HARK master.

I'm going to merge this PR and then make separate issues for these outstanding bugs.