Refactor `num_cpu` across the board

[valentin-krasontovitsch]

When it comes to figuring out the config objects, this is related to #3957, #3918, #4004, and more generally #3986.

Documentation of num cpu is covered by #1379, and a user reported bug may be found in #3979

These are my findings regarding num cpu -

found in	source	used by
queue config	config file / dict NUM_CPU	enkf main; submission of jobs
ecl config	data file DATA_FILE	enkf main
enkf main	queue config, or ecl config	simulator and run models
subst config	config file / dict NUM_CPU, or data file DATA_FILE	substitution list

To me it seems like enkfmain (or res config, if you will, wherever) could handle all getting of the num cpu from different sources (config file / dict and data file), and pass the value to the different config objects as needed - and based on this usage list, at least ecl config would no longer need to hold / know about num cpu.

[oyvindeide]

There is a test in: https://github.com/equinor/ert/pull/4017 which tests how NUM_CPU is set when no NUM_CPU is set in the config, but DATA_FILE is set, this seems to be relevant for this issue.

[valentin-krasontovitsch]

thanks for pointing it out, i'm gonna check it out!

[valentin-krasontovitsch]

indeed we are aware of the mechanism tested for there, and have no intention of changing anything about that logic - so no problem!

at least if i understood correctly, the logic for among others substitution config is: if num cpu is given in the user config, use that value, otherwise fetch it from the data file if a data file is provided.

[valentin-krasontovitsch]

turns out we can completely remove num cpu from the queue config (which is obvious and natural, really) - it's literally not used.