Figure out exactly which scenario / problem to work on.

[abmarnie]

Jeff and Justin worked on different fluid inference problems, some of which are implemented in BayesianShape.

My plan is to pick a single scenario, and then work from the "top down" to re-implement it, naively inlining everything and using common and well-established packages to handle lower level things (such as linear algebra). My goal isn't to make something modular, it is to make something:

1. Reproducible
2. Testable
3. Straightforward to understand

Once that is done, I might look into performance. Performance tips are here and here; the TLDR is that every expression RHS has to be written so that the LHS is completely inferable, and heap allocations need to be avoided (which is something I am used to thinking about due to working in C#).