Produce name-sorted CRAM

[vinjana]

Do not discard the alignment information and produce a name-sorted CRAM. The alignment information will allow for high reference-based compression.

The alignment and the sequence information anyway is separated in the SAM format (CIGAR strings + position + strand information vs. separate sequence column).

Secondary and supplementary alignment should be discarded (optionally), to have better compression.

SAM attributes might also be discarded, to safe more space.

The result would be just sequences and alignments in order to maximize compression level.