Progressively render the exercise Rmd to accurately capture prep and result environments

[gadenbuie]

Fixing envir_prep and envir_result

This PR fixes rstudio/gradethis#199, namely we use a two-step rendering process to ensure that we accurately return the prep environment in envir_prep.

Previously the envir_prep environment contained the results of evaluating the global setup chunk, but not the exercise-specific setup chunks. With the two step evaluation via rmarkdown::render(), we now guarantee that envir_prep contains the results of the global setup and the exercise setup.

We also ensure that envir_result is a cousin environment duplicated from envir_prep. If an error occurs in the setup code, envir_result and envir_prep are identical. If an error occurs in the user code, envir_result reflects evaluated code up to the error.

New Testing Infrastructure 🏗️

This PR also adds exercise mocking functions that let us create mock exercises without having to write and evaluate a tutorial Rmd. This simplifies testing and lets us directly test render_exercise() and evaluate_exercise(). I focused mostly on testing render_exercise(), so we'll likely want to increase testing on evaluate_exercise() in the near future.

Exercise Checker for Debugging 🚫 🐛

As part of this testing infrastructure, I added a new internal function debug_exercise_checker(). It serves as a counterpart to debug_event_recorder() and may be helpful to tutorial authors who are trying to write custom error checking functions.

Here's a small reprex. By setting the default exercise.checker to debug_exercise_checker, Submit Answer returns the values of the arguments expected by a custom exercise checking function.

---
title: "A tutorial"
output: learnr::tutorial
runtime: shiny_prerendered
---

```{r setup, include=FALSE}
library(learnr)

tutorial_options(exercise.checker = learnr:::debug_exercise_checker)
```

```{r example-setup}
x <- 1
```

```{r example, exercise=TRUE}
# user submits
x <- 4
x - 1
```

```{r example-solution}
x + 3
```

```{r example-check}
# check code
```

PR task list:

• Update NEWS
• Add tests (if possible)
• Update documentation with devtools::document()

[nischalshrestha]

Looking good! Just had a comment + question.

[schloerke]

LGTM pending the learnr::: in the test is removed