Scipion-pyworkflow

- Improving subset creation
- PLUGIN_MODULE variable can be used in host.conf to customize mpirun command: mpirun.PLUGIN_MODULE --> mpirun.relion)
This approach implies having this pattern for all mpi plugins that use mpi.
- Alternatively, in the scipion.conf file a variable called PARALLEL_COMMAND_XXX could be defined for a more
versatile way to have a mpi command only for some plugins. Example:
PARALLEL_COMMAND_RELION=mpirun.relion -np %%(JOB_NODES)d --map-by node %%(COMMAND)s
NOTE: in this config file, %_ (valid and needed in host.conf) should be replaced by %%
- GPU Ids are automatically anonymized (0-len(GPUs)) when using queues
- Bug-fix: creation time is not loaded when loadig run.db (updating a protocol)
developers:
- Adding a Streaming section in ProtStreamingBase
- runJob call without %(GPU)s will not reserve GPU slots anymore.

V3.10.4: hotfix: Parallelization with GPUs is fixed and step queue executor fixed with the new implementation.

V3.10.2: hotfix: Executor frees the GPU/s used via step id instead of node.

developers:
- stepsExecutionMode = STEPS_SERIAL at class level. Should be a class attribute for future versions. Now both approaches are valid.