`augment.tune_results()` doesn't always return tibbles

``` r
library(tidymodels)

lr_spec <- parsnip::logistic_reg() %>% parsnip::set_engine("glm")

set.seed(1)
two_class_dat <- as.data.frame(two_class_dat)
bt1 <- rsample::bootstraps(two_class_dat, times = 30)

set.seed(1)
fit_1 <- fit_resamples(
  lr_spec,
  Class ~ .,
  bt1,
  control = control_resamples(save_pred = TRUE)
)
augment(fit_1) %>% class()
#> [1] "data.frame"
```