Throw an error for @importFrom unknown export

[MichaelChirico]

Closes #1409.

This is a rough / off-the-cuff version. Wanted to checkpoint here to get high-level feedback, there's still TODO:

• Adapt to internal package style (e.g. cli)
• Add tests
• Add NEWS

[hadley]

This seems reasonable to me.

[MichaelChirico]

I can't make snapshots from my current machine, so I'll try and pull them from the GHA artifacts.

Should be ready for review anyway.

[hadley]

Thanks!

[MichaelChirico]

two issues with this PR in retrospect as surfaced by #1570:

1. importing can be like "%>%" or `%>%` or '%>%' but getNamespaceExports() will always just be the plain symbol (%>%)
2. importing[!unknown_idx] might drop everything --> output becomes importFrom($PKG,)

How would you recommend approaching these?

1. Is a bit tricky because we have to normalize the quotes, I guess we can do gsub("^(`|'|\")(.*)\\1$", "\\1", importing). This could be a helper alongside has_quotes() Initially, I was worried about collision where you try and import a really weirdly-named object like as.name("'%>%'"), but I guess {roxygen2} has always made that difficult. And a quick search doesn't turn up any usage of this anyway: https://github.com/search?q=lang%3AR+%2F%5Cn%5Cs*%28%22%27.*%27%22%7C%27%22.*%22%27%7C%60%22.*%22%60%7C%60%27.*%27%60%7C%22%60.*%60%22%7C%27%60.*%60%27%29%5Cs*%3C-%2F+-path%3A.Rd&type=code
2. Should this method just return(NULL) if all(unknown_idx)?

[hadley]

1. Yeah, I think strip_quotes() seems reasonable.
2. Yes, I think so.

[MichaelChirico]

thanks! that's basically how I did #1574 so let's continue there.