(follow up to #52)

For convenience, we could have a flag for the fastest and slowest compression levels.

They must conflict with each other and the newly added --level.

Implementation details: you can use i16::MAX and i16::MIN because levels are always clamped to the min-max range.