Make it visually easier to recognize similar sub-patterns

[lukaslueg]

Moved over from lukaslueg/macro_railroad#12

---

In many macros, multiple branches share a lot of similar sub-patterns. It would be easier to visually read the diagrams if the similar sub-patterns were aligned. E.g. aligning the 4 occurrences of macro_name ! into one column in this example, and also aligning the 2 occurrences of values = vf ,.

And for macros where a sub-patterns appears verbatim in multiple branches of the macro (or multiple times in one branch), it would make sense to highlight that they are not just similar but the same, by factoring out that sub-pattern into its own sub-rule diagram (and then inserting the name of the sub-rule (or an arrow to the sub-rule diagram) in place of the occurrences of the sub-pattern).

[lukaslueg]

In both cases the underlying problem is probably (related to) the frequent subgraph mining problem. gSpan is the only thing that comes to mind for me.

I'm unsure about the alignment: We currently completely circumvent the problem of having primitives know about their surrounding and alignment rules. That is, we don't solve the optimization problem of respecting as many alignment rules as possible while simultaneously not blowing up the diagram. A place where this shows up is deep Stacks which return to their exit-point (see example #12) , which could be aligned downwards instead of returning upwards.
To solve this, an implementation similar to CSS-positioning of absolute and relative may suffice.

If we had an FSM-implementation, extracting / rewriting macros into multiple diagrams should be easy.