apply knitr print method for inline r code

[maxheld83]

This replaces the custom roxygen2 evaluation of inline code with default knitr inline.
As such, this enables knit_print() methods and closes #1179.

I'll list the potential problems with this first:

1. there might have been a good reason to go with a custom treatment of inline code, that I am ignorant of.
2. this change may change existing documentation (so it's kind of a "breaking" change):
   1. knit_print() methods are applied (which might not always work with the roxygen2 markdown engine)
   2. in the custom engine `r x <- 10` would yield [1] 10.
      Knitr (and console) behavior is to print nothing in this case.
      I've changed the tests accordingly, but this might bite some (hopefully very few?) people.
      Since the whole dynamic R was marked as a beta feature, perhaps this minor breaking change is acceptable.

Advantages include:

1. brings dynamic code inside roxygen2 more in line with rmarkdown/knitr expectations.
2. more knitr features (knit_print() and some other niceties.
3. overall somewhat streamlined test, usage and docs
4. minor change may be acceptable because feature was marked as experimental.

To the extend that they pertain to inline dynamic R code, I've also streamlined the tests and documentation:

• examples are simpler (but largely useless)
• there are a couple more tests

Remaining todos before this read for review:

• some real-life testing, general maturation
• docs
• tests
• proofreading

[hadley]

Do you want to try and get this into this release? I'm aiming for release around the end of next week.

[hadley]

We could finish this one off too, if you wanted?

[maxheld83]

I'm happy with this now, I think it's ready for review.

I have:

• merged in main and resolved all merge conflicts
• fixed build failures (an old release didn't support on.exit(after = FALSE), so I used the superior withr idiom instead).
• completed another proofreading run

@hadley you've previously r-lib/testthat#1308 recommended not to be too clever about reusing code between tests and documentation, because it makes the tests harder to read.
I've violated and am reusing code between the vignette and the tests because:

• the pattern is already in use with several tests/testthat/roxygen-block-*.Rs
• the necessary duplication really would be quite onerous, and things are nice and small like this.

Does that seem ok to you @hadley?

[hadley]

Thanks for working on this! I do think it's better to make the tests self-contained even if it leads to some repetition. See my explanation below.

[maxheld83]

@hadley thanks for the thorough review -- very instructional.
rlang::local_bindings() is a great tip, and no, I would've never looked for it in rlang.

Hope I addressed everything ok and the tests still pass 🙏.

[hadley]

Thanks!