Add support for hanging commas

[trestletech]

Closes #128

Uses rlang::dots_list instead of list to support trailing commas with missing arguments a la

tagList("hi", )
div("one", )

I'll confess that I was guessing at which of the expectations in test-deps were going to turn out to be lists or dot_lists. That makes me wary that I might have missed something or that this may be an incomplete migration.

This does change the hash for singletons, but I added a test to confirm that the old hash will still work.

Testing notes

Add a hanging comma at the end of some tags when defining a Shiny UI and it should work e.g.

tagList("hi", )
div("one", )

Multiple trailing commas should throw an error:

div("one", , )

Non-trailing commas should throw an error:

div( ,"one", )

[wch]

The format is usually:

Fixed #128: .... (#135)

I'll just make the change directly.

[hadley]

I’d suggest using list2() here, and I don’t think you should need to change the tests.

[jcheng5]

(sorry just noticed this PR now)

This gives HTML tag functions tidyeval semantics now, doesn't it? Is there any situation where that could break existing code? (Maybe where an expression being passed to an HTML tag includes dplyr calls that need to interpret the !! themselves?)

Even if there are such cases, depending on how edgy they might be, it might not be a showstopper anyway. But the fear of such breaking changes was what stopped me from immediately making this change, hoping we might find or make a variant of list2 that didn't do tidyeval--or else convince ourselves that the edge cases are vanishingly edgy.

[hadley]

Not general tidy evaluation semantics, just tidy dots semantics (i.e. you can use !!!). (i.e. I think it's very unlikely that this will change semantics in practice; switching from list() to list2() is something that we generally do without worry)

[jcheng5]

Thanks @hadley that's perfect