Parallelise "filter reads" step in Makefile

[lweasel]

At the moment, the filter reads step calls a Bash script, "filter_reads", which then calls the "filter_sample_reads" python script for each sample, to do species separation on the mapped reads for that sample.

One way of changing this might be to:

• alter "filter_reads", so that for each sample, it first splits the mapped reads BAM file for each species into multiple (roughly equally sized?) name-ordered BAM files (in a temporary directory). It may be possible to do this with the --filter option of "sambamba view"? (Although that introduces a dependency on sambamba). Actually, this is probably going to be the hardest bit of the process to implement.
• call a python script which will use the subprocess module to launch multiple instances of the "filter_sample_reads" python script for each chunk of the input BAM files. Poll these processes to work out when they've all finished.
• concatenate the filtered BAM files for each input chunk, for each species. Tidy up the temporary directory.

[s-heron]

The block splitting has been implemented. On a test sample (1A1) it ran in 13m30.560s to split it into 4 blocks for both species. Several minutes can be shaved off this through termination of the sambamba file stream after each start id has been extracted, but I haven't found a workable way to do this. Parallelised execution of filtering on the blocks will be implemented after the new filter script has been written.

[s-heron]

Wrote & commited the parallelisation control script; filter_control.py