implement prefix aggregators

[jeisner]

I really find myself wanting prefix aggregators already, so that I don't have to introduce too many new names. This is particuarly true for queries, where they're just throwaways anyway:

query max= count(Word).   % most frequent word

Here's a fancy example involving a nested pair of aggregators:

query argmax= [(+= 1 for brown(Sentence,Position)), Sentence].  
      % length and ID of longest sentence

This assumes that the prefix form of argmax= returns a pair, as in the design suggested in this https://github.com/nwf/dyna/issues/29#issuecomment-20404238 comment.