break zoomark's dependency on pool_threadlocal, zoomark_orms dependency on engine-bound session

[Mike Bayer]

investigate why zoomark requires pool_threadlocal. it seems to be using autocommit so there shouldn't be issues turning it off but I was getting data errors, etc. I hardly looked at it at all so might be something simple.

zoomark_orm also seems to require that the session is bound to the engine directly instead of pulling it from metadata. Might be something simple as well.

[jek]

I can't replicate that here w/ PG 8.1 linux or 8.3 linux/osx. Reverting the 3 engine/pool and session line changes from dbda75ec5f98061b87cf616b468232ebc68a9133 passes completely here, ignoring out of date call count assertions.

[Mike Bayer]

OK i cant reproduce either now.

[Mike Bayer]

Removing milestone: 0.5.xx (automated comment)