S(q,w) calculations are slow compared to MD with GPU

**Description of the error**
If we use MD with GPU, most of the compute time is spent running the S(q,w) calculations. Here are some basic results with my laptop, which I obtained by adding some print statements in a few places.

Intel Core Ultra 7 155H
RTX 4050

argon-openmm.py
MD CUDA, MDMC ("OMP_NUM_THREADS" = "4")
Equilibration: 2.25s
Production: 6.23s
S(q,w) calc: 153.3s

argon-openmm-mdanse.py
MD CUDA, MDMC ("OMP_NUM_THREADS" = "4")
Equilibration: 1s
Production: 1.5s
S(q,w) calc: 278.3s

water-openmm.py
MD CUDA, MDMC ("OMP_NUM_THREADS" = "4")
Equilibration: 6.89s
Production: 533.4s
S(q,w) calc: 22242.9s

water-openmm-mdanse.py
MD CUDA, MDMC ("OMP_NUM_THREADS" = "4")
Equilibration: 8.1s
Production: 183.5s
S(q,w) calc: 1445.3s