Population size is scaling with ex post heterogeneous distribution approximation sizes

[sbenthall]

Using the WHITESHARK population now, i.e. with this:

$ python run_any_simulation.py --population WHITESHARK --quarters 1 ../output/20230320_s

Creates a population of 600 AgentTypes with AgentCount 1.

Why are there 600 agents? Because this is the number of points in the discretized space of heterogeneity over which we compute the HARK solutions:

whiteshark_approx_params = {"CRRA": 3, "DiscFac": 2, "RiskyAvg": 10, "RiskyStd": 10}

But this is odd. It makes sense (sort of) for the population to scale with the number of ex ante heterogeneity gridpoints (though this feels hacky even in this case...), but it doesn't make sense for the population to scale with the ex post heterogeneity gridpoints.

Related to #195 in so far as this involves the population generation logic. Assigning to @alanlujan91

[sbenthall]

Push this to 4.2 milestone as it's not necessary for GG Shark

[sbenthall]

Punt because for SPARK the consumer population will be homogeneous.