Deprecate the optional_aes field in Geom/Stat

[teunbrand]

This PR aims to fix #6393.

It takes the 'violent' path of deprecating the field.
We can also choose to delay deprecating the field.

[thomasp85]

I'm slightly unsure why we need this. Yes, it "simplifies" the class definition, but on the other hand it also removes some self-documenting aspects of it.

Do you have other reasons for this beside "we can do without"?

[teunbrand]

The main reason I'd say is that it is a cause of confusion when learning how to build ggplot2 extensions because it isn't really clear what should go into optional_aes. It is better now that #6426 describes the field, but it remains somewhat redundant and befuddles a clear path for how to implement optional aesthetics.