Fix image name collision in docs

[ffreyer]

Description

Fixes #4388

Type of change

• Bug fix (non-breaking change which fixes an issue)

[jkrumbiegel]

I think the fix should rather be to hash the code after the transformation step a line below?

I actually don't remember the reason for this design, but I wonder if one couldn't just use the page plus a running index to disambiguate. Using the transformed code there would still be collision problems when two examples have the same code, like a simple fig where some state is modified in between to show an effect on the visual output.

[ffreyer]

The transformation step also adds save($(id)...) so the id needs to exist beforehand

[jkrumbiegel]

Ah ok missed that on my phone, can you try the running index version instead, where the hash is page, number_of_figure_block? Not sure how to get the number, probably needs some kind of state sadly. But we already have state, maybe just reuse that. Could also just do number_of_figure_globally as we have a dict of all of them I guess?