metaquast does not scale for large number of references

[nick-youngblut]

This is mainly to warn users:

Running metaQUAST (QUAST 5.0.2) with >500-1000 bacterial genome references & >1 million Illumina reads takes many days to finish, even when using many threads (eg., >7 days to finished for 2000 ref genomes with 2 mil metagenome reads, with 12 threads).

This lack of scaling seems to mainly be due to running QUAST on each ref genome separately. I tried to split metaquast into 2 separate steps: 1) everything up to the per-ref QUAST runs 2) the per-ref QUAST runs and all the final sections. This would allow for running the per-ref QUAST runs as separate cluster jobs (eg., 1000's of QUAST jobs in parallel). However, the metaquast.py is so intertwined that creating module sections of the code would likely require a full re-write. One major issue is the fact that qconfig cannot be pickled. Instead of explicit arg passing to functions, qconfig is used all over the place, which causes all kinds of headaches.