RethinkDB internals: Patching for fun and throughput

This feels like one of those optimizations that makes sense now, but very quickly just becomes bloat. We've seen a lot of heroic optimizations that make sense at the time and improve performance, which only a few years later actually slow things down. Remove the "optimization" and go with the simpler code, and things become much faster.

Optimizing for spinning hard disks seems like one of those optimizations that will soon come back to bite you. Graph out the Iops improvement of SSD's. We're getting to the point where we can think of an SSD as secondary memory and treat hard disks as offline storage.