step_dummy with no variable selected

@kjell-stattenacity  In cases where `step_dummy()` receives _no variables_, it should not just fail. 

```r
library(tidymodels)

zdat <- tibble(
  y = c(1, 2, 3), 
  x = c("a", "a", "a"),
  z = 3:1
  )

transform_recipe2 <- recipe(y ~ x, data = zdat) %>%
  step_zv(all_predictors()) %>% 
  step_dummy(all_predictors()) %>%
  prep(training = zdat)
```