Suggestion for new documentation

[marcpaterno]

The inner workings of Phlex are complex. Diagrams and associated descriptions can help with understanding the complexity, if they are kept up to date. There is already one such helpful diagram in phlex/core/multilayer_join_usage.dot, shown below.

This diagram was produced in DOT; I would propose we stick with DOT for additional diagrams to avoid needlessly multiplying the number of tools we need to have at our disposal when working with the code. Inside Visual Studio Code, the Graphviz Interactive Preview plugin is one way to view DOT diagrams.

I propose that it would be good to have a series of such diagrams, showing this level of implementation detail for several simple configurations. Each diagram should have a single purpose, such as showing the behavior of an unfold node. The picture below shows a hand-drawn diagram that does this:

This diagram (produced by Kyle) was extremely valuable while reviewing the PR #571 that implemented some upgrades to make it actually work.

As a starting set, I propose diagrams that show the details of workflows with the following configuration of nodes:

1. a provider that produces an index set family that goes to a single transformer. This is nearly the simplest graph possible.
2. a provider that returns an index set family that goes to an unfold, which produces products in a new layer, followed by a transform acting on the products from the new layer.
   3.a provider that returns an index set family, going into a fold over that index set family to yield some summary product.

I suspect that more complicated diagrams might eventually be useful, but these three will help illustrate the implementation of many of the important innards of the system.