Test on r-devel

[lionel-]

Closes #1066

To track upstream breaking changes to critical components of our infrastructure:

• The way we integrate with ReadConsole (Take charge of parsing and evaluating inputs #960) and more generally the frontend API
• Graphics devices API

Second commit fixes sensitivity of a test on HTML produced by R:

        FAIL [   0.834s] (296/861) ark lsp::help::tests::test_help_from_function
  stdout ───

    running 1 test
    test lsp::help::tests::test_help_from_function ... FAILED

    failures:

    failures:
        lsp::help::tests::test_help_from_function

    test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 490 filtered out; finished in 0.82s

  stderr ───

    thread 'lsp::help::tests::test_help_from_function' (22703) panicked at crates/ark/src/lsp/help.rs:384:38:
    called `Option::unwrap()` on a `None` value

[DavisVaughan]

Do you happen to know what the change in R was?

Did you look at what this is connected to and try to interactively cause whatever LSP feature it is connected to to trigger in R-devel and R-release after this change? I just really want to make sure we don't accidentally regress on something here, and I'm not sure we have the best coverage at the moment.

[lionel-]

R -q -e 'cat(tools::Rd2HTML(utils:::.getHelpFile(help("match"))))' 2>/dev/null | grep '<h3' 

On R 4.5:

<h3>Description</h3>
<h3>Usage</h3>
<h3>Arguments</h3>
<h3>Details</h3>
<h3>Value</h3>
<h3>References</h3>
<h3>See Also</h3>
<h3>Examples</h3>

On R 4.6:

<h3 id='_sec_description'>Description</h3>
<h3 id='_sec_usage'>Usage</h3>
<h3 id='_sec_arguments'>Arguments</h3>
<h3 id='_sec_details'>Details</h3>
<h3 id='_sec_value'>Value</h3>
<h3 id='_sec_references'>References</h3>
<h3 id='_sec_seealso'>See Also</h3>
<h3 id='_sec_examples'>Examples</h3>