Incorporate recordPlot functionality for recording and replaying plots as objects - #686

Fixes #121

By adding grDevices::recordPlot() to the very end of the main tinyplot() function, we gain the ability to record (assign) plots as objects and then replay them later.

pkgload::load_all("~/Documents/Projects/tinyplot/")
#> ℹ Loading tinyplot

a = plt(Sepal.Length ~ Petal.Length | Species, data = iris)

plt(1) # some other random plot

a  # replay the saved plot

Here's a little plt_add gotcha, though...

plt_add(type = "lm") # gotcha: the last plot "call" was plt(1)!

But note that nothing stops us from layering immediately... and we can record and replay from a later layer too. Here, demonstrating with a theme to drive home the point.

plt(Sepal.Length ~ Petal.Length | Species, data = iris, theme = "socviz")
b = plt_add(type = "lm")

plt(1) # random placeholder again

b # replay the saved, layered plot object

Note: Due to the "canvas" nature of base graphics, we can't suppress the initial display. Unless, that is, we do something tricksy like write to a dummy device / file on disk and then recall later.

png("dummyplot.png")
dev.control("enable") # default for png is "inhibit" -> disallows recording
p = plt(Nile, type = "hist")
dev.off()
#> png 
#>   2
p # replay directly on our main, interactie device

^{Created on 2026-08-20 with reprex v2.1.1}

Discussion: To be clear, I'm not convinced that we want to add this functionality. But some quick testing, at least, suggests that it doesn't add any real overhead. Apart from the plt_add() gotcha that I highlight above, I also haven't hit any unintended side effects. Still, I'd like others to kick the tyres too, before we consider whether merging is a good idea.