Evaluate cost of removing quasiquotation support

[hadley]

Since it's not that useful, not commonly used, and makes it harder to test functions that use unquoting.

[hadley]

Judging by https://github.com/search?q=org%3Acran+%2F%21%21%5B%5E%21%5D%2F+path%3Atests%2Ftestthat%2F&type=code&p=2, there's enough usage it's not going to be worth eliminating. Not to mention, a couple of hours of work in testthat itself, since we'd need to rework the condition expectation helpers in order to capture labelled values earlier (although it's arguable we should do that anyway?)