gpu template matching should queue subvolumes on single gpu

[McHaillet]

Splitting a tomogram into subvolumes for template matching should be able to queue the subvolumes sequentially on single (or multiple gpu's). Currently the number of subvolumes needs to equal the number of cores/gpu's provided. Specifying more subvolumes than cores leads to an error.

[McHaillet]

This would also be solved if we switch to importing SBC-Utrecht/pytom-template-matching-gpu