unexpected discrepencies with read_shocks in FramedAgent example

[sbenthall]

Originally discussed here: https://github.com/econ-ark/HARK/pull/1071#issuecomment-972894786

In the FramedAgent example, I've tried to use the read_shocks functionality to use the exact same shocks in the PortfolioConsumerType and the PortfolioConsumerFrameType, on the theory that the simulation histories should then be identical.

My current use as of #1071 of this isn't working, in that the two simulated shock histories are divergent.

On the other hand, making this change does make the simulation histories look more similar.

Likely there are two issues buried in here: one about proper use of read_shocks, which could be a better documented workflow, and another with the alignment of the simulation mechanics of the framed and unframed models.