Support subdataframes

Currently SubDataFrames are not iterable tables, which is unfortunate as it's actually surprisingly clumsy to convert them to DataFrames. The current method to iterate on DataFrames could probably have a weaker signature and maybe that'd be enough to have it work on any AbstractDataFrame. The main application would be to be able to use it in a by call, see this discourse post

julia> using IterableTables

julia> df = DataFrame(x = rand(10));

julia> by(TableTraits.isiterabletable, df, :x)
10×2 DataFrames.DataFrame
│ Row │ x        │ x1    │
├─────┼──────────┼───────┤
│ 1   │ 0.67292  │ false │
│ 2   │ 0.921996 │ false │
│ 3   │ 0.517816 │ false │
│ 4   │ 0.822496 │ false │
│ 5   │ 0.120007 │ false │
│ 6   │ 0.289319 │ false │
│ 7   │ 0.113851 │ false │
│ 8   │ 0.826421 │ false │
│ 9   │ 0.881944 │ false │
│ 10  │ 0.279972 │ false │

Start with the current DataFrames iteration method and the TableTraits.isiterabletable entry point mentioned in the issue. Check how SubDataFrames behave when passed to by, and compare that with the existing DataFrame behavior. Done means SubDataFrames can be used as iterable tables in the demonstrated by call.