Memory constraints for concat AOFlagging on many-core systems

The low core count for the aoflagger step resulted in many jobs launching at once on the 96-core machine I run on. The default memory percentage of 15% that is passed caused the pipeline to crash for me as combined memory usage (75 GB * 16 jobs) exceeded the available 512 GB.

As a test I tried changing the CWL file to also include a memory requirement to

  - class: ResourceRequirement
    coresMin: 6
    ramMin: 125000

giving it 125 GB (to err on the safe side based on the 15% default setting), which seems to have fixed the issue for me.

It is unclear to me whether CWL allows a percentage of RAM to be specified, but in case not perhaps instead of aoflagger.memoryperc we could explore linking aoflagger.memorymax and ramMin?