Load during build preparation is proportional to number of project build results

Found while addressing CONTINUUM-1621, due to the way that Continuum find the previous build results during build preparation, memory and CPU load increases proportionally to the number of build results for a project. This will eventually result in 100% cpu utilization and OutOfMemoryErrors.

To produce this, I imported a project and replicated a set of build results until there were over 1,000,000. When I restarted Continuum, the CPU and memory utilization (viewable using jconsole) would spike until it threw and OOME.

By producing a heap dump and analyzing it, I was able to pinpoint the problem area, a bad query in the DAO.