Small fix in Perlmutter GPU sbatch script (#5683)

Changes in Perlmutter GPU job script:
from `#SBATCH --cpus-per-task=16` to `#SBATCH --cpus-per-task=32`.

This is to request (v)cores in consecutive blocks.

GPU 3 is closest to CPU cores 0-15, 64-79, 
GPU 2 to CPU cores 16-31, 80-95, 
...

If `--cpus-per-task=16`, MPI ranks 0 and 1 are mapped to cores 0 and 8.
If `--cpus-per-task=32`, MPI ranks 0 and 1 are mapped to cores 0 and 16.

Visual representation

![pm_gpu_vcores_mpi](https://github.com/user-attachments/assets/edf0721f-7321-49ab-bf37-4b55a7c422cc)

---------

Co-authored-by: Axel Huebl <[email protected]>

Tools/machines/perlmutter-nersc/perlmutter_gpu.sbatch

@@ -17,7 +17,7 @@
 # A100 80GB (256 nodes)
 #S BATCH -C gpu&hbm80g
 #SBATCH --exclusive
-#SBATCH --cpus-per-task=16
+#SBATCH --cpus-per-task=32
 # ideally single:1, but NERSC cgroups issue
 #SBATCH --gpu-bind=none
 #SBATCH --ntasks-per-node=4
@@ -34,7 +34,7 @@ export MPICH_OFI_NIC_POLICY=GPU
 
 # threads for OpenMP and threaded compressors per MPI rank
 #   note: 16 avoids hyperthreading (32 virtual cores, 16 physical)
-export OMP_NUM_THREADS=${SLURM_CPUS_PER_TASK}
+export OMP_NUM_THREADS=16
 
 # GPU-aware MPI optimizations
 GPU_AWARE_MPI="amrex.use_gpu_aware_mpi=1"