Remove duplicate functionality from pv211_utils.evaluation_metrics, add ETA, and prevent the pickling of systems

[Witiko]

Since #7, we've had evaluation metrics in the pv211_utils.evaluation_metrics module. Here is a number of issues with the implementation:

• Both calc_map and mean_average_precision implement the same functionality, see also a related comment from the review of Implement the MUNI/33/0769/2022 R&D project #8. Let's remove one of them and update the library to always use the single remaining implementation?
• Both calc_map and mean_average_precision pickle the system when num_workers > 1, which may easily fail or slow the evaluation down significantly when the system references a large object that is expensive to pickle.
  • For calc_map, this seems to be a bug, since the code does the heavy lifting of saving the system to a global and forking only to pickle the system anyways, see also a related comment from the review of Implement the MUNI/33/0769/2022 R&D project #8.
  • For mean_average_precision, there is not even a pretense of avoiding pickling the system, see also a related comment from the review of Implement the MUNI/33/0769/2022 R&D project #8.
• The evaluation can take a bit. Let's be friendly to users and give them an ETA? See also a related comment from the review of Implement the MUNI/33/0769/2022 R&D project #8.

After fixing the above, let's make num_workers=None the default? Pool(processes=None) uses all the available CPUs.