FEAT: add Mermaid graph rendering

[Flomber]

Closes #159

[Flomber]

Found issue with rendering long lines eg. in graph edges.

isospin_magnitude ∊ [0, 1/2, 1, 3/2 ]

gets displayed as

isospin_magnitude ∊ [0, 1/2, 1, 3/2

[redeboer]

Good job! 💯
Some smaller comments below.

One general thought I had: it seems that Mermaid (as opposed to Graphviz) supports LaTeX, even if not for all backends. Rewriting the example of #159:

```mermaid
graph LR
    A["$$J/\psi$$"] --> N0[ ]
    N0 --> N1["$$f_0(980)$$"]
    N0 --> 0[$$\gamma$$]
    N1 --> 1[$$\pi^0$$]
    N1 --> 2[$$\pi^0$$]
    style A fill:#FFFFFF, stroke:#FFFFFF;
    style N0 fill:#FFFFFF, stroke:#FFFFFF;
    style N1 fill:#FFFFFF, stroke:#FFFFFF;
    style 0 fill:#FFFFFF, stroke:#FFFFFF;
    style 1 fill:#FFFFFF, stroke:#FFFFFF;
    style 2 fill:#FFFFFF, stroke:#FFFFFF;
```

Maybe worth posting a follow-up issue. Similarly to #348, one could select LaTeX as rendering with a flag, for instance.

Rendering on Sphinx

Rendering on VS Code Jupyter

Rendering on Jupyter Lab

Rendering on GitHub ❌

graph LR
    A["$$J/\psi$$"] --> N0[ ]
    N0 --> N1["$$f_0(980)$$"]
    N0 --> 0[$$\gamma$$]
    N1 --> 1[$$\pi^0$$]
    N1 --> 2[$$\pi^0$$]
    style A fill:#FFFFFF, stroke:#FFFFFF;
    style N0 fill:#FFFFFF, stroke:#FFFFFF;
    style N1 fill:#FFFFFF, stroke:#FFFFFF;
    style 0 fill:#FFFFFF, stroke:#FFFFFF;
    style 1 fill:#FFFFFF, stroke:#FFFFFF;
    style 2 fill:#FFFFFF, stroke:#FFFFFF;

Loading

[redeboer]

Better to declare this class with @attrs.define to reduce boilerplate code, just like GraphPrinter, which is declared with @define(on_setattr=_check_booleans).

[Flomber]

Does this definition satisfies this?
Also AI mentioned, that in the definition of the GraphvizPrinter class has the weakness, that it also accepts invalid constructor combinations.
Should this also be fixed? A shared attrs validator was a recommendation.

[redeboer]

• This returns a str, and line 134 uses it directly as sorted(..., key=..., reverse=True). This makes priority 9 sort above 10, and "NA" outranks every numeric priority. The rule list in the rendered label ends up in the wrong order whenever any priority reaches double
  digits.
  Returning an int with an explicit sentinel for NA (-1, or float("-inf") if negative priorities are possible) would fix it.
• Minor problem: in the same regex, the \- does not need escaping outside a character class.

Note that this exact same function exists under _dot.py!

[Flomber]

Should be fixed, but please review... I'm not exactly sure if the fix is what you asked for 🙃