Simplify usage of prepared tx statements

It allocated all transaction-specific prepared statements when the transaction is created, simplifying the places where such statements are used, and improving performance, as it's executed only once per transaction, instead of per operation.

As a result, it removes tons of boilerplate code. That explains the code coverage shrinking :-)

When applied to the tracking package, I observed a considerable performance improvement (importing a large log directory took from 168 to 159 seconds).