coord_munch can convert closed shapes to closed form

[teunbrand]

This PR aims to fix #3271.

In brief:

• It repeats the first row in a group to after the last row in a group.
• Due to this, the rect_to_poly() function in geom_rect() made a redundant edge. While amending this, I found a more efficient way to polygonise rectangles, which is 2x faster for n = 2 rectangles, but 200x faster for n = 1000 rectangles.
• A few visual tests changed due to closing the shape, but I manually checked them and they're invisible changes.

A visual comparison:
Before:

library(ggplot2)

offset <- function(value, offset) as.vector(outer(value, offset, `+`))

df <- data.frame(
  x = offset(c(-0.5, 0, 1, 1.5, 1, 0, 0, 0, 1, 1), c(0, 2)),
  y = offset(c(0.5, 0, 0, 0.5, 1, 1, 0.25, 0.75, 0.75, 0.25), c(0, 2)),
  group = rep(1:2, c(10, 10)),
  subgroup = rep(1:4, c(6, 4, 6, 4))
)

ggplot(df, aes(x, y, group = group, fill = factor(group))) +
  geom_polygon(aes(subgroup = subgroup)) +
  coord_polar()

After:

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

ggplot(df, aes(x, y, group = group, fill = factor(group))) +
  geom_polygon(aes(subgroup = subgroup)) +
  coord_polar()

^{Created on 2022-12-03 by the reprex package (v2.0.1)}

[teunbrand]

The snapshots of maps have revert to their old state after removing the closing point.
However, rectangles in non-linear coordinates lose a closing point. Linejoins on rectangles look like they should.

[thomasp85]

Yes - rectangles always had one too many so this change is considered a fix

[thomasp85]

LGTM, but if you could provide a test that shows how this fixes rendering of closed paths it would be great