Refactor ReadsSparkSource to accommodate non-sorting bam output

[jamesemery]

Currently PrintReadsSpark only supports coordinate sorted input files because ReadsSparkSource automatically sorts the output. This is not necessary as many spark tools are written in such a way that they won't affect the sorting order of the underlying read inputs. For these tools we should provide the ability to write the contents of the bam using the original partitioning order.