Track ggplot2 version in plot

[teunbrand]

This PR aims to fix #6714.

Whenever a plot is created, the new version property is populated with the current version.
If a plot is to be rendered, an error is thrown if the plot's version is not the same as the current version.
This should typically occur when a plot has been saved to disk with a prior version of ggplot2, and now no longer works.

There are a few open questions:

• Should we do this at all?
• Should we throw a warning instead of an error?

We can also do a revdepcheck to weed out all the packages that misbehave.

[aphalo]

@teunbrand Hi Teun! This is an interesting question. If you decide to implement this, I suggest using a warning, unless you are sure all old plots will not render at all. Sometimes, it can be useful to get some idea of what a plot is about, even if inperfectly rendered. Another thought is that it would be better to use "object versions" than 'ggplot2' versions. Why issue an error for a change from version 4.0.1 to 4.0.2? I would change the label only when there is a significant compatibility break. In such a case, the warning (or error), could give the 'ggplot2' versions that can render the object as a range, which would be a lot friendlier to users in the case they cannot rebuild the object.
I (almost) never save ggplot objects to disk. So, as you mentioned in #6714, the main solution is to discourage users from saving these objects to disk. However, nothing prevents users from saving objects or expecting this to work, say sharing a ggplot object with somone who does not have the code to regenearte it. Another possible situation is a user working interactively at the command prompt, saving the R workspace, and expecting to be able to render the objects in the workspace at a later time, possibly after an update to 'ggplot2'. So I think that a coarser grained test, triggering an error only when rendering is not possible would be more helpful. (I am assuming rendering is possible at least across minor versions.)

[teunbrand]

Thanks for your thoughts Pedro! I get that this would be easier from a user perspective, but from a maintenance point of view it'd become tedious to test and verify if there is between-version compatibility and doing the bookkeeping. It is not even guaranteed for minor version updates. People can share plots with themselves and oneanother as data + code, or as graphics, but shouldn't do this as ggplot objects.