convert expressions for levels

[topepo]

Related to #145

The calibration objects store factor levels as expressions and we are seeing errors when subsetting columns or setting class levels in new factors:

library(probably)
#> 
#> Attaching package: 'probably'
#> The following objects are masked from 'package:base':
#> 
#>     as.factor, as.ordered
data(two_class_example, package = "modeldata")
w_calibration <- cal_estimate_logistic(two_class_example,
                                       truth = truth,
                                       estimate = c(Class1, Class2))

cal_apply(two_class_example, w_calibration, pred_class = Class)
#> Error in `purrr::map()`:
#> ℹ In index: 1.
#> Caused by error in `[<-.data.frame`:
#> ! anyNA() applied to non-(list or vector) of type 'symbol'

^{Created on 2024-09-20 with reprex v2.1.0}

[simonpcouch]

Awesome—this does the trick for tidymodels/tailor@probabilities-46. I'll move forward with enabling tests and sending that PR in once this is merged!