es-search-references: Encapsulating object subgraphs needs a built-in "self learning" function.

I built [my `OneToOneStrongMap` class](https://github.com/ajvincent/es-membrane/blob/ab78dab3776e0ea487d08088de1bcb616de86a21/projects/v2-es-membrane/stage_utilities/source/collections/OneToOneStrongMap.ts) atop `WeakMap` and [`WeakStrongMap`](https://github.com/ajvincent/es-membrane/blob/ab78dab3776e0ea487d08088de1bcb616de86a21/projects/v2-es-membrane/stage_utilities/source/collections/WeakStrongMap.ts). However, the object graphs I generate in `es-search-references` partly strip that away, showing the internal guts of the latter while building out the former.

Since all `WeakStrongMap` instances have the same shape, we could compact a subgraph into a much smaller shape. For the `OneToOneStrongMap`, showing that "I have the map and one weak value as enough to reference another weak value" without the rest of the noise would be pretty valuable in compacting the graph.

How to do that is going to require some careful thought.