This allows us to have faster fuzz cycles since the fuzzer is in-process.
With the current setup it successfully finds the (fixed) DOCTYPE buffer overrun in ~50 minutes (on a single core).