F() term: allow the filtering statistic to have multiple elements.

[krivit]

For example,

nw~F(~edges, ~nodefactor("a", levels = TRUE))

could return a statistic for every level of attribute a. Currently, the way to do that is something like

nw ~ For(l = sort(unique(nw%v%"a")),
         F(~edges, ~nodefactor("a", levels = I(l)))
         )

In principle, this could be implemented exactly as shown there, but there would need to be a way to autodetect the levels---or perhaps us Sum() to subset the filtering term.