RankSelect uses double the space needed

[jorgehermo9]

Currently, the RankSelect data structure stores rank samples for every s bits, but it stores the superblocks_1 samples and the superblocks_0 samples. It is not necessary to store the superblocks_0 samples, because given any position i, its associated rank_1 sample is in the position i/s on the superblocks_1 sample vector, and the superblock_0 sample at that point can be easily calculated as i - superblocks_1[i/s] without needing to store it...
Storing also the superblocks_0 when it could be easily calculated, duplicates the space used by this data structure.