rip off collections.namedtuple for Query so that it can be pickled

[Mike Bayer]

For 2.4 and 2.5 support, use the source to Py2.6's "collections.namedtuple", which execs a class from string source thus allowing pickle-ability.

questions:

1. do we use namedtuple's API verbatim ? or modify it for SQLA purposes (such as the "keys" attribute??)
2. is this version of namedtuple much slower than the quickie method we are doing ?

[Mike Bayer]

collections.namedtuple is not even pickleable. in 0.6 we are using a new tuple that just has a module-level class.

[Mike Bayer]

Removing milestone: 0.6.0 (automated comment)