Enable distribution vectors in tibbles

[zeileis]

Alex @alexpghayes, I recently noticed that we had provided enough S3 methods for distribution objects that they could be included in data.frame objects. However, it was not possible to include them in tibble objects.

Now I was able to work out (with the help of Hannah @hfrick, thanks!) that we only need vec_proxy() and vec_restore() methods in order to use distribution vectors in tibble. The vec_proxy() adds the data.frame class (i.e., the internal data structure underlying distribution objects) and vec_restore() removes it again. For example, you can get predicted distributions in tibbles:

library("distributions3")
library("tibble")
data("FIFA2018", package = "distributions3")
m <- glm(goals ~ difference, data = FIFA2018, family = poisson)
nt <- tibble(difference = -1:1)
nt$dist <- prodist(m, newdata = nt)
nt
## # A tibble: 3 × 2
##   difference dist                    
##        <int> <Poisson>               
## 1         -1 Poisson(lambda = 0.8181)
## 2          0 Poisson(lambda = 1.2370)
## 3          1 Poisson(lambda = 1.8704)