Automatic checkpoint states on every line

I was wondering what the original motivation for the [`CheckpointManager`](https://github.com/3b1b/manim/blob/master/manimlib/scene/scene_embed.py#L162) was since it forces the users to split the scene into small chunks that start with a comment. This is really nice as it gives you a Jupyter-like experience and we fully hopped onto that train with [Manim Notebook](https://github.com/Manim-Notebook/manim-notebook) by introducing the concept of Manim Cells into VSCode.

But just as a figment, imagine what a more interactive preview experience could look like. Imagine a [**Manim Animation Debugger (MAD)**](https://github.com/Manim-Notebook/manim-notebook/issues/120), where users can step through the `construct()` method just like they would on regular code and preview the scene line by line, or even run until the next breakpoint. And while doing so, see variable states evolve in a debug pane. The code would become a bit like a vertical video timeline that users can easily navigate. They could jump to any line and start previewing there, without the need for an extra comment at the start.

So as a first step to this vision, I was wondering if **checkpoints could be automatically created for every single line of the animation and not just for comment lines**. Are there any memory or performance concerns in doing so, in particular for very long scenes? Maybe I'm missing something entirely.

The CheckpointManager is in manimlib/scene/scene_embed.py. First, understand how it currently creates checkpoints at comment lines. Then, explore the scene rendering flow to see where line-by-line checkpoints could be inserted. Consider memory implications for long scenes by profiling checkpoint state size. A prototype would modify the checkpoint logic to trigger on every executable line in construct().

Scritto dal modello di indicizzazione a partire dal testo della issue.