wishlist: caching http object retriever

[Edd Dumbill <edd usefulinc com>]

Quite a few backends do HTTP fetches.  We want to cache these objects so
we're not going back to the web all the time (especially under
cluechaining circumstances where we might get a clue three times, we
don't want to do three fetches.)

A class that took a URL and handed back a stream, doing caching where
necessary would rock.  It should respect expiry times too, either set in
the HTTP headers or from the caller.

-- Edd