cast-volume: avoid loading the full volume in memory
It looks like we should faster than expected modify the code to read one slice at the time when computing min / max and casting slices...
It looks like we should faster than expected modify the code to read one slice at the time when computing min / max and casting slices...