Update stacks column names for new tune

[topepo]

Closes #248 and closes #249

The new results:

library(tidymodels)
library(stacks)
#> Registered S3 method overwritten by 'butcher':
#>   method                 from    
#>   as.character.dev_topic generics

data("penguins")

# change the species so that it has a - in it
dat <- penguins %>%
  mutate(species = paste0(species, "-1"))

# tune a glmnet model w 3 penalties
tuned <- recipe(species ~ bill_length_mm + bill_depth_mm +
                  flipper_length_mm, data = dat) %>%
  step_impute_mean(all_numeric_predictors()) %>%
  workflow(multinom_reg(engine = "glmnet", penalty = tune())) %>%
  tune_grid(dat %>% vfold_cv(3),
            metrics = metric_set(mn_log_loss),
            grid = crossing(penalty = c(.0001, .001, .01)),
            control = control_stack_grid())

data_st <- stacks() %>%
  add_candidates(tuned)

colnames(data_st)
#>  [1] "species"                               
#>  [2] ".pred_Adelie.1_tunedpre0_mod1_post0"   
#>  [3] ".pred_Adelie.1_tunedpre0_mod2_post0"   
#>  [4] ".pred_Adelie.1_tunedpre0_mod3_post0"   
#>  [5] ".pred_Chinstrap.1_tunedpre0_mod1_post0"
#>  [6] ".pred_Chinstrap.1_tunedpre0_mod2_post0"
#>  [7] ".pred_Chinstrap.1_tunedpre0_mod3_post0"
#>  [8] ".pred_Gentoo.1_tunedpre0_mod1_post0"   
#>  [9] ".pred_Gentoo.1_tunedpre0_mod2_post0"   
#> [10] ".pred_Gentoo.1_tunedpre0_mod3_post0"

^{Created on 2025-08-01 with reprex v2.1.1}

[simonpcouch]

Based on the PR description, it seems like you're writing that stacks "just works" with the new config ids?

I do see this error in the CRAN R-CMD-check:

══ Failed ══════════════════════════════════════════════════════════════════════
── 1. Failure ('test-stacks-columns.R:43:3'): stacks can accommodate outcome lev
".pred_Adelie.1_tunedpre0_mod2_post0" %in% colnames(data_st) is not TRUE

`actual`:   FALSE
`expected`: TRUE 

And a good few in the GH check.

[EmilHvitfeldt]

The error that simon found has been addressed in 61c62ec