make compute_metrics_reloaded.py faster

[naga-karthik]

When computing metrics using ANIMA and MetricsReloaded one can quickly notice that ANIMA is much faster probably because it is written in C++ whereas MetricsReloaded is pure Python. we gain better readability (python vs c++ code) at the cost of speed.

discussing with @valosekj, our wrapper script could be made faster by computing the metrics in parallel instead of a simple for as it is done now.

one idea: use multiprocessing ie spawn multiple workers on multiple subjects (just like the -jobs arg in sct_run_batch ?