do not pass empty vector to Tables.columntable

[bkamins]

@nalimilan - following your comment on Slack this is a DataFrames.jl fix. In it I propose even more extreme approach - do not generate any columns if the returned vector is empty. This will make consistent behavior of DataFrame, and GroupedDataFrame, as it is now a corner case when they differ.

The behavior before this PR is:

julia> df = DataFrame(a=[1,2,3])
3×1 DataFrame
 Row │ a
     │ Int64
─────┼───────
   1 │     1
   2 │     2
   3 │     3

julia> gdf = groupby(df, :a)
GroupedDataFrame with 3 groups based on key: a
First Group (1 row): a = 1
 Row │ a
     │ Int64
─────┼───────
   1 │     1
⋮
Last Group (1 row): a = 3
 Row │ a
     │ Int64
─────┼───────
   1 │     3

julia> combine(df, :a => (a -> [(b=1,c=2)][1:0]) => AsTable) # we try to identify columns
0×2 DataFrame
 Row │ b      c
     │ Int64  Int64
─────┴──────────────

julia> combine(gdf, :a => (a -> [(b=1,c=2)][1:0]) => AsTable) # we just ignore empty groups
0×1 DataFrame
 Row │ a
     │ Int64
─────┴───────

TODO: tests/news - after we decide what we want

[bkamins]

@nalimilan - this should be good to review

[bkamins]

Thank you!