added maxforks to gpu queues

[Mitchob]

Summary

Queue	Route max_queued	Execution max_run [u:PBS_GENERIC]	Execution max_run [p:PBS_GENERIC]	max_run_res.ncpus [p:PBS_GENERIC]
gpuvolta	1000	Unrestricted	Unrestricted	1920 CPUs
dgxa100	50	1 job/user	1 job/project	—
gpuhopper	50	10 jobs/user	8 jobs/project	720 CPUs

Add maxForks caps to GPU-queued processes in proteinfold.config and bindflow.config

proteinfold.config

• RUN_ALPHAFOLD2_PRED|RUN_ALPHAFOLD2|COLABFOLD_BATCH (gpuhopper): maxForks = 2
• RUN_BOLTZ (gpuhopper): maxForks = 2

bindflow.config

• BINDCRAFT (dynamic queue dgxa100/gpuvolta via params.use_dgxa100): maxForks = { params.use_dgxa100 ? 2 : 8 }, matching the tighter dgxa100 per-user run limit vs. the looser, contention-bound gpuvolta.

Not changed

• wisps.config (RUN_BOLTZ, COLABFOLD_BATCH on gpuvolta) left as-is for now since only using gpuvolta with no real limits.

[tlitfin]

• Could you please also change wisps and bindflow to use gpuhopper?
• Could you also add maxforks to wisps config?
• Could you also update proteindj config? There will be multiple gpu-bound processes in that workflow so may need a global queuesize or similar.

[Mitchob]

Things worth double checking:

• wisps if/else statement, do we want it removed so it is just gpuhopper or is switching volta > hopper ok?
• global executor set to to for proteindj. Is thats what you meant by your last bullet point?

[eileen-xue]

Does BINDCRAFT need 24 hours?

Other things look good to me.