VK: Split 05-recruits into four Cardinal Virtue tutorials = - #16

…th predictions(datagrid())

plot_predictions() on a parsnip logistic_reg() fit with a continuous
`condition` exhausts memory under marginaleffects 0.32.0, failing the
tutorial render. Exercises 10-11 now sweep MonthlyIncome via an explicit
datagrid() passed to predictions() -- equivalent output, and datagrid()
is already introduced in Exercise 6.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XAfJeVM4b87jrAjQeQUkZt

…plot columns resolve

datagrid() with a function-valued grid variable did not carry MonthlyIncome
into the predictions() output on the CI marginaleffects, breaking geom_ribbon()
with "object 'MonthlyIncome' not found". Build the grid as an explicit
tibble()/expand_grid() instead -- predictions(newdata = <df>) always returns
the newdata columns, so the plot code resolves on every marginaleffects version.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XAfJeVM4b87jrAjQeQUkZt

…tic predictions-12

plot_predictions(mod_logistic, by = "OverTime", type = "prob", draw = FALSE)
crashes under marginaleffects 0.32.0. Exercise 12 now pipes
avg_predictions(by = "OverTime") into ggplot(). With Exercises 10-11 already
switched to predictions(), plot_predictions() is no longer used in the section,
so the covered-functions lists and recap prose are updated to match.

…ctions call

avg_predictions(mod_logistic, by = "OverTime", type = "prob") does not echo
the OverTime column on the CI marginaleffects, breaking aes(x = OverTime) with
"object 'OverTime' not found". Run avg_predictions() on extract_fit_engine()
(the underlying glm) instead: response scale is already P(leaving), no type or
group filtering, and by= reliably returns the OverTime column.

…ine()

predictions()/avg_predictions() on the parsnip logistic_reg() wrapper do not
reliably echo the grid or by= variable as a named column under the CI
marginaleffects, so ggplot(aes(x = MonthlyIncome / OverTime)) fails with
"object not found". Exercises 5, 10, 11, and 12 now call these functions on
extract_fit_engine(mod_logistic) (the underlying glm, type = "response"),
which returns those columns and the probability scale directly -- no type =
"prob" or group filter. Printed chunks (1-4, 6-9) keep the tidymodels wrapper.