Use COPY instead of INSERT during massive sync

This commit probably shouldn't be merged as-is. I'm still baffled by the collection of classes and templates used to get data into the database, so I'm sure this code isn't using them for their intended purpose. But someone who understands the code may be able to take some ideas from it and fit it into the existing framework.

This reduces the time for a full replay (up to p2p sync) from 61253 sec to 54660, a little over 10% speedup.