Make .check_censor_model() error for non-censored-regression models

[hfrick]

closes #971

Regression and classification models also have $censor_probs even though it's just an empty list, so this check function didn't error before.

Test is to be updated in extratests.

With this PR:

library(parsnip)
library(survival)

lung2 <- lung %>%
  dplyr::mutate(surv = Surv(time, status), .keep = "unused")

not_a_censor_model <- fit(linear_reg(), mpg ~ ., data = mtcars)

# this now errors
parsnip:::.check_censor_model(not_a_censor_model)
#> Error in `parsnip:::.check_censor_model()`:
#> ! The model needs to be for mode 'censored regression', not for mode
#>   'regression'.

# so also this errors now with the better message
.censoring_weights_graf(not_a_censor_model, lung2)
#> Error in `.check_censor_model()` at parsnip/R/survival-censoring-weights.R:226:2:
#> ! The model needs to be for mode 'censored regression', not for mode
#>   'regression'.

^{Created on 2023-05-18 with reprex v2.0.2}

[hfrick]

An alternative is to make changes in fit() and fit_xy() to only fit the additional model for censoring probabilities if the main model has mode censored regression. On the one hand, the model_fit objects then only carry what they need. On the other hand, the model_fit objects look somewhat different on what the mode is, was there hesitation around type-stability there? Do you have a preference for either of the two options?

[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.