Add /forceSave workaround for 1.9's AsyncChunkLoader silliness.
/allAtOnce now implies /forceSave
Make the memory limit a little more conservative to ensure we don't hit GC overhead errors on low memory servers
Once the memory limit has been hit, wait for ram to decrease below the limit by at least 10% before resuming
Tweak the NextTickList bug workaround - instead of flushing the list, just ensure it stays below a threshold.
Should fix the minor lag spikes from the fix.