Fix naming in df_stats() when parens are used.

Since implementing this by wrapping () with `I`, we get some unfortunate naming in `df_stats()`.
```r
df_stats( ~ (height > 65), data = Galton, mean)
##   mean_I(height > 65)
## 1           0.6269488
```