[Redfoot-general] First cut at BerkeleyDB triple store

[<aku...@me...>]

See http://www.amk.ca/files/python/rdfdb.py for a persistent triple
storage implemented on top of BerkeleyDB using the bsdb3 module.
A simple test case is included at the bottom of the file.

Questions:

* Should I make it match the interface of triple.py exactly?  Right
now it doesn't match, but fixing that shouldn't be difficult.

* It would be nice to have a variant of search() that uses 2.1/2.2's
generators to avoid constructing a list of all matching tuples.
That wouldn't match triple.py, unless the feature was added there,
too.

I haven't been able to figure out how to integrate it with ZODB's
transaction machinery yet, but will continue to look into that.

--amk                                                             (www.amk.ca)
Our duty, if not our conscience, is clear.
    -- President Borusa, in "Arc of Infinity"