feat: Rework view() to better work with RStudio and Positron

[DavisVaughan]

Closes #1551 (while I was here, I may as well...)

In RStudio and Positron, when View(foo) is called, we check if foo exists as a symbol bound in the parent environment. If it does, we can "track" the lifetime of that object and auto update the data viewer when the user makes changes to it. This doesn't work with view()

Screen.Recording.2024-09-30.at.10.56.25.AM.mov

This is because in view() we were actually inlining the whole data frame object, rather than the expression the user typed in. So RStudio and Positron can't find anything to "track". This was also causing issues in Positron too actually (the inlined data frame could overwhelm us) posit-dev/positron#4702.

@lionel- and I thought about this a bit and came up with an approach where we capture the user input to view() and use that to reconstruct an equivalent call to View() that we call in the parent environment. This mimics the idea of the user calling View(foo) directly in the global environment, so that RStudio/Positron can "track" foo correctly.

• In the case that x isn't a data frame, we don't do any of this. We "make" x into a data frame ourselves, so RStudio and Positron won't have anything to "track" in these cases so we don't do anything special
• In the case that x is a data frame but comes from something like view(cbind(foo, foo)), we do the fancy thing to recall this as View() in the parent frame, which works fine, but of course RStudio and Positron still won't have anything to track because this is an ephemeral object. This is expected.

It works like this now:

Screen.Recording.2024-09-30.at.10.54.40.AM.mov

Screen.Recording.2024-09-30.at.10.55.18.AM.mov

---

Side note, how do you add news bullets in tibble?

[krlmlr]

Thanks, great!

NEWS bullets are added by CI/CD from the PR title once merged, don't worry about it.

[krlmlr]

Thanks! Happy to release as needed.