Reproducible simNGS runs and multi-library check for Fastq_screen

[brainstorm]

• Fastq_screen was not checking all-to-all bloom filters and organisms, now it does, which fills some gaps when comparing with FACS.
• SimNGS will now produce the same dataset over runs (checked with md5sum). It should solve the debate on reproducibility about this "random" component of the testsuite.
• Revert to single-ended reads, default simNGS behaviour. This is the way we are testing fastq_screen right now due to the lack of paired-end support in FACS, @inodb ;)
• Fixed JSON to reflect threads on fastq_screen. Also updated current organism being tested.