Don't use multiple worker threads for integration

It appears this triggers some weird bug when used in combination with
bwa-mem2, which causes the alignment rule to crash randomly.
6 jobs for speedup in 98 minutes and 7 seconds (queued for 1 second)