Eliminate duplicate warning on joins

[hadley]

df1 <- tibble(a = 1:2)
df2 <- tibble(
  b = structure(1:2, foo = "bar"),
)

out <- left_join(df1, df2, by = c("a" = "b"))
#> Warning messages:
#> 1: Column `b`/`a` has different attributes on LHS and RHS of join 
#> 2: Column `a`/`b` has different attributes on LHS and RHS of join 

@krlmlr could you take a look at this? I'm not familiar enough with the join code to figure out what's happening, but it looks like DualVector is constructed twice, once with left and right flipped.