`step_collapse_cart()` gives uninformative results when using `outcome` wrongly

library(embed)
library(modeldata)

# no outcome - bad error
recipe(Street ~., data = ames) |>
  step_collapse_cart(Neighborhood) |>
  prep() 
#> Error in `step_collapse_cart()`:
#> Caused by error in `purrr::map2()`:
#> ℹ In index: 1.
#> ℹ With name: Neighborhood.
#> Caused by error in `training[[y_name]]`:
#> ! Can't extract column with `y_name`.
#> ✖ Subscript `y_name` must be size 1, not 0.

# multiple outcomes - bad error
recipe(Street ~., data = ames) |>
  step_collapse_cart(Neighborhood, outcome = vars(MS_SubClass, MS_Zoning)) |>
  prep() 
#> Error in `step_collapse_cart()`:
#> Caused by error in `purrr::map2()`:
#> ℹ In index: 1.
#> ℹ With name: Neighborhood.
#> Caused by error in `training[[y_name]]`:
#> ! Can't extract column with `y_name`.
#> ✖ Subscript `y_name` must be size 1, not 2.

# factor outcome - disregarded
recipe(Street ~., data = ames) |>
  step_collapse_cart(Neighborhood, outcome = vars(Street)) |>
  prep() 
#> 
#> ── Recipe ──────────────────────────────────────────────────────────────────────
#> 
#> ── Inputs
#> Number of variables by role
#> outcome:    1
#> predictor: 73
#> 
#> ── Training information
#> Training data contained 2930 data points and no incomplete rows.
#> 
#> ── Operations
#> • Collapsing factor levels using CART: <none> | Trained

# numeric outcome - expected
recipe(Street ~., data = ames) |>
  step_collapse_cart(Neighborhood, outcome = vars(Sale_Price)) |>
  prep() 
#> 
#> ── Recipe ──────────────────────────────────────────────────────────────────────
#> 
#> ── Inputs
#> Number of variables by role
#> outcome:    1
#> predictor: 73
#> 
#> ── Training information
#> Training data contained 2930 data points and no incomplete rows.
#> 
#> ── Operations
#> • Collapsing factor levels using CART: Neighborhood | Trained

^{Created on 2024-09-11 with reprex v2.1.1}

Start by reproducing the four step_collapse_cart() examples from the issue, covering no outcome, multiple outcomes, a factor outcome, and a numeric outcome. Read the implementation and existing tests for step_collapse_cart() to identify where outcome validation occurs. Done means invalid outcome usage produces an informative result while the numeric-outcome example retains its current behavior.