Parallelize sampling

[jk-rieck]

This could be done by using dask_bags or multiprocessing to work on different tracks in parallel. But is complicated because we need to increment i. (see sample.sample, the for ed in np.arange(0, len(tracks)):-loop)