Reorganize evol ops

[felixhekhorn]

After #194 we realized that the content of evolution_operator needs some reorganizations.

• there are many elements that are quite old, e.g. related to the old runcards and runner structure
• there is a certain asymmetry between OME and evolution operators, mainly historically generated
• eko.member should be a part of the subpackage
• after First jets implementation #242 we could consider dropping multiprocessing
  • because for single $\mu^2$ values we're fast enough, and it will be possible to compute multiple ones in parallel
  • because of the many drawbacks of multiprocessing in Python (including the memory duplication)
  • even just to simplify, since we'll only need to parallelize at a single level, and There should be one-- and preferably only one --obvious way to do it.
• and just generically outdated code