Reduce disk space contention on CI runner VMs created by gsrm
As a corrective action for production#1976 (closed), reduce the number of jobs serially handled by each runner VM from 30 to 10.
Each time a runner VM runs a job, it accumulates residual disk space used by that docker container's volumes. Sometimes an unlucky VM runs a combination of jobs that fills its disk. This has started happening fairly often, and it is starting to affect the efficiency of development work.
For more background, see summary notes here: production#1976 (comment 329467200)
This is the cheapest of the easy options. If it turns out to be insufficient, we can revisit the other options.