Add method that compresses the chunks with lowest uncompressed state_group ids

[Azrenbeth]

This is an alternative to #65 which compressed the rooms with the most uncompressed data.

This way won't work on the largest rooms first, so it might take longer to get the same space savings, but it doesn't need to count the number of rows in state_groups_state which would have taken a long time if a room hadn't been compressed before and had a lot of state.