fix: multiprocessing in generating crystal graphs

[bsplu]

The flag multiprocessing = True can cause the computer to run out of resources.

This is because:
The num_threads keyword in kernel.py sets the torch threads.
But the process_worker uses multiprocessing, which conflicts with the torch threads.

Solution:
I modified the multiprocessing flag from bool to int and I set the torch threads to one in data.py. I restored them after the process_worker finished.

Suggestion:
To optimize performance and memory, I suggest using multiprocessing = 0 and changing num_threads to set the threads. This avoids large memory consumption when creating crystal graphs.