Removing our datasource/pooling implementation.

Having received another report[1] of the lack of robustness of our pooling
implementation I think we should scrap our datasource and pooling
implementation. I previously advocated keeping it around because it
"basically worked" and didn't really cost us anything to keep it. Now
we're aware that it doesn't really work and I for one don't want to spend
time fixing it when there are better options out there.

I spent some time today testing jakarta's dbcp[2] and I couldn't find
anything our code does that it cannot and there are plenty of additional
features. Dynamic pool sizing, removing broken connections, and even
statement pooling are available. I was impressed.

Would anyone like to make a case for keeping our implementation around?

Kris Jurka

[1] http://gborg.postgresql.org/project/pgjdbc/bugs/bugupdate.php?1109
[2] http://jakarta.apache.org/commons/dbcp/