Automatic limit setting in scale_*_binned leads to values being silently coerced to NA

When making a plot with scale_color_binned in 4.0.2 I discovered that when breaks are set manually, values may be displayed as NA even if they are within the bins specified. In the reprex below I make a scatter plot of hp vs mpg in the built-in mtcars dataset, with points colored by displacement. When we just set scale_color_binned with no options, everything looks fine:

If we set manual breaks of (0, 100, 150, 200, 500), the rightmost point (which has displacement of 71.1) is now shown as gray (i.e. NA), despite being within the scale breaks:

If we explicitly set the scale limits to (0, 500) the problem is solved:

Reprex:

library(ggplot2)

data(mtcars)

min(mtcars$disp)
#> [1] 71.1

# this makes a scatter plot of hp vs mpg, with points colored by
# displacement; this works
ggplot(mtcars, aes(x=mpg, y=hp, color=disp)) +
    geom_point() +
    scale_color_binned()

# If we set breaks manually, the rightmost point is colored gray for
# NA even though it is within the breaks.
ggplot(mtcars, aes(x=mpg, y=hp, color=disp)) +
    geom_point() +
    scale_color_binned(breaks = c(0, 100, 150, 200, 500))

# explicitly setting limits solves the problem
ggplot(mtcars, aes(x=mpg, y=hp, color=disp)) +
    geom_point() +
    scale_color_binned(breaks = c(0, 100, 150, 200, 500),
        limits=c(0, 500))

Start by running the supplied mtcars reprex with scale_color_binned() and with manual breaks, then compare it with the explicit limits case. Trace the automatic limit-setting path for scale_color_binned and add a regression test for the 71.1 displacement value; done means values inside the supplied breaks are not coerced to NA.