I'd like this key to be called computed rather than _computed please, as leading underscores mean "private" in Python.

Also, we need an entry in the schema for this new key, so that the Python codegen adds the property accessors and so it appears in the reference documentation etc etc.

Overall this looks good, but I'm seeing some odd behaviour with a figure like this one: https://codepen.io/nicolaskruchten/pen/wvWBmyQ

here the long x label is causing some automargin edge case: the right margin is increased because of the slanted label, but the bottom margin would exceed the cutoff so it resets. This is a known problem with automargins, and we don't need to tackle it now BUT the computed margin still shows r: 80 for me, which is incorrect in this case.

Overall this looks good, but I'm seeing some odd behaviour with a figure like this one: https://codepen.io/nicolaskruchten/pen/wvWBmyQ

here the long x label is causing some automargin edge case: the right margin is increased because of the slanted label, but the bottom margin would exceed the cutoff so it resets. This is a known problem with automargins, and we don't need to tackle it now BUT the computed margin still shows r: 80 for me, which is incorrect in this case.

Here is a minimal codepen to illustrate the bug you mentioned.
@nicolaskruchten could you please confirm that it "sometimes" happening depending on the length of the long tick label?

I haven't poked too much at when exactly it happens, sorry. Can we fix it in the case that I identified?

Given that the issue already exists in fullLayout._size I'd be happy to consider that an independent bug and merge this PR as is, for its own part this PR looks great. But @nicolaskruchten I'd understand if you want to delay merging this until we solve that, since the bug isn't currently exposed to users.

Let's merge this as-is for release in 1.57 for our sponsor, and then open a separate issue for the _size and prioritize fixing that one, as now some internal things are external ;)

💃 merging then...