add butcher methods for dbarts::bart()

[simonpcouch]

Related to #234. I haven't added a NEWS entry, as I anticipate putting in a few of these PRs today, and don't want to introduce merge conflicts for you!

Unfortunately, the "heaviest" entry by far, $fit, is required for prediction.😫 That slot is only occupied when keeptrees = TRUE, which is not a default.

library(butcher)
library(dbarts)

fit_bart <- function() {
  boop <- runif(1e6)
  dbarts::bart(mtcars[,2:5], mtcars[,1], verbose = FALSE, keeptrees = TRUE)
}

res <- fit_bart()
res_b <- butcher(res)

weigh(res)
#> # A tibble: 9 × 2
#>   object              size
#>   <chr>              <dbl>
#> 1 fit             1.85    
#> 2 yhat.train      0.256   
#> 3 varcount        0.0167  
#> 4 sigma           0.00805 
#> 5 call            0.00196 
#> 6 first.sigma     0.000848
#> 7 yhat.train.mean 0.000304
#> 8 y               0.000304
#> 9 sigest          0.000056
weigh(res_b)
#> # A tibble: 9 × 2
#>   object              size
#>   <chr>              <dbl>
#> 1 fit             1.85    
#> 2 first.sigma     0.000848
#> 3 y               0.000304
#> 4 call            0.000112
#> 5 sigest          0.000056
#> 6 sigma           0.000048
#> 7 yhat.train      0.000048
#> 8 yhat.train.mean 0.000048
#> 9 varcount        0.000048

^{Created on 2022-12-22 with reprex v2.0.2}

[juliasilge]

This looks great! Thank you so much 🙌

[github-actions]

This pull request has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.