Support nesting()

[mgirlich]

Fixes #238.

nesting() needs to be treated separately from other variables in expand() and the resulting data.table code quickly gets a bit ugly, in particular as there is no nice support for a cross join by data.table.

TODO

• add tests
• handle factors
• add news
• check if row order is the same as for dataframe

factors

When used with factors, expand() uses the full set of levels, not just those that appear in the data.

It is possible to handle factors the same way but not so nice because the type of a column is usually not known (respectively only when it wasn't mutated). I see two options to have the same behaviour for data.table as for dataframes:

• use levels(x) %||% x for every column. This is quite ugly
• define a special function and add it in front of the generated data.table code

[markfairbanks]

@mgirlich should this one be converted to a draft/work-in-progress? Did you run into some issues when implementing?