Improve on unpack() column name duplication errors

[DavisVaughan]

Closes #1425 (Unless we decide otherwise)
Related to #1367

I'd say that this resolves #1425 by giving a much better error message.

It also resolves this comment #1367 (comment), again with a much better error message.

It does not resolve the original issue presented in #1367 about using some automatic separator for unnamed elements. I think that is related but can be thought about separately. I've kept that one open.

In general this improves errors for unpack(), unnest(), unnest_wider(), and the separate_wider_*() family.

---

There are two types of name duplication issues that can occur when unpacking, which I have called:

• inner / inner duplication (unpacking multiple columns that have duplicate inner names)
• outer / inner duplication (unpacking 1 column that duplicates an outer name)

This PR attempts to give better error messages for both of those cases when names_sep = NULL and names_repair = "check_unique", i.e. the user has not attempted to solve these issues themselves yet.

I tried not to mention "unpacking" in the error because these bubble up from other functions.

library(tidyr)

# inner / inner
df <- tibble(
  x = list(tibble(a = 1)),
  y = list(tibble(a = 2))
)
unnest(df, c(x, y))
#> Error in `unpack()` at tidyr/R/unnest.R:181:2:
#> ! Can't duplicate names across the modified columns.
#> ✖ These names are duplicated:
#>   ℹ `a`, across `x` and `y`.
#> ℹ Use `names_sep` to disambiguate using the column name.
#> ℹ Or use `names_repair` to specify a repair strategy.

# outer / inner
df <- tibble(
  x = 1, 
  y = list(tibble(x = 2))
)
unnest(df, y)
#> Error in `unpack()` at tidyr/R/unnest.R:181:2:
#> ! Can't duplicate names between the modified columns and the original data.
#> ✖ These names are duplicated:
#>   ℹ `x`, from `y`.
#> ℹ Use `names_sep` to disambiguate using the column name.
#> ℹ Or use `names_repair` to specify a repair strategy.

You would get more indented bullets if there are additional problems (see tests)