postgresql storage should not be using serializable transactions

[jack]

This is not banking software. I believe that the original author must have used serializable because he was unfamiliar with transaction levels. We should be using the normal read committed level, which is enough for the purpose at hand. After all, there is no retry code if a serializable transaction fails is there? So it is implemented incorrectly anyway.

Also, we need to get rid of the overhead of setting the transaction isolation level for every transaction. This can be done when the connection is established and will save a lot of time.

Also, we should not be using BEGIN/END blocks on singleton inserts.

We're working on paches for all these, but I wanted to open a bug here for feedback and to track the status.

[smoku]

I'm very open to the patches.

[smoku]

There is no reason for patches really.

Just comment-out <transactions/> in sm.xml and you're done.