Running large jobs during training/workshops

[natefoo]

Right now, jobs during trainings can go to LS6 (specifically, the vm-small partition) where:

1. only 4 jobs can run at a time
2. only 16 jobs can be queued total before additional submissions are rejected with error
3. there may be considerable queue wait.

On usegalaxy.org for trainings we use TIaaS and drastically decrease memory/core resources so that tools run on normal nodes, but this isn't always possible for some tools, like Kraken2 which load the entire selected database into memory.

We could use reservations for this on LS6, but:

1. I don't know if the reservations would be whole nodes or consumable resources
2. I don't know if slurm-drmaa supports --reservation (but could be added if it doesn't)