WIP: Extracting guide data

[teunbrand]

This PR aims to address issues such as #5004.

Briefly:
It provides a new function, guide_data() to programmatically access values and labels (internally called 'keys') of axes and legends.

Less brief:
Currently, the way to reconstruct, for example axis tick positions, is to rummage around in ggplot2's structures. As laid out by Thomas here, querying data structures is a gray area. Providing a convenient accessor to the guide information is nice for two reasons. 1) It provides a user with a reliable way to extract the information and 2), as Thomas mentions in the blog, it leaves the developers free to change internal structures without worrying too much that a downstream dependency breaks due to internal rearrangements. It should, in theory, prevent surprises such as #5063, #5029 or #3436 (comment).

WIP:

• I had to refactor the ticks code for CoordSf a bit, by capturing the logic in a function that I could reuse in guide_data(). However, should the graticule-to-ticks logic be a ggproto method of CoordSf or should it live as a separate function?
• Is guide_data() a good name? Are there better ones? I'll take suggestions.
• Any other feedback is welcome too!

The examples reprex'ed:

devtools::load_all("~/packages/ggplot2/")
#> ℹ Loading ggplot2

# A typical plot
p <- ggplot(mtcars) +
  aes(mpg, disp, colour = drat, size = drat) +
  geom_point() +
  facet_wrap(vars(cyl), scales = "free_x")

# Getting guide data for position aesthetic in particular panel
guide_data(p, "x", i = 1, j = 2)
#> [[1]]
#>           x .value .label y
#> 1 0.0959596     18     18 0
#> 2 0.3484848     19     19 0
#> 3 0.6010101     20     20 0
#> 4 0.8535354     21     21 0

# Getting guide data for a legend
guide_data(p, "size")
#> [[1]]
#>       size .label
#> 1 2.662822    3.0
#> 2 3.919819    3.5
#> 3 4.779645    4.0
#> 4 5.477285    4.5

# If guides are merged, `guide_data()` reports merged guide data
guide_data(p + guides(colour = "legend"), "size")
#> [[1]]
#>    colour .label     size
#> 1 #1A3855    3.0 2.662822
#> 2 #28557C    3.5 3.919819
#> 3 #3874A5    4.0 4.779645
#> 4 #4894D0    4.5 5.477285

^{Created on 2022-12-08 by the reprex package (v2.0.1)}

[teunbrand]

Closing this PR in favour of #5337.