Parallelization refactor adding joblib prefer="processes"

[vitorandreazza]

This PR is a refactor to higher parallelization by utilizing the prefer="processes" option of the joblib.Parallel class. This change permits the full usage of the CPU cores and processing power, significantly reducing the total runtime of the .fit.
For example, in this simple test:

series = pd.Series([random.randint(1, 1000) for _ in range(10000)])
f = Fitter(series, timeout=20)
f.fit(progress=True)

Current running time output:

Fitting 113 distributions:   0%|          | 0/113 [00:00<?, ?it/s]
Fitting 113 distributions:   2%|▏         | 2/113 [00:00<00:10, 10.80it/s]
Fitting 113 distributions:   5%|▌         | 6/113 [00:00<00:08, 12.40it/s]
...
Fitting 113 distributions:  96%|█████████▋| 109/113 [01:17<00:03,  1.23it/s]
Fitting 113 distributions:  97%|█████████▋| 110/113 [01:17<00:01,  1.55it/s]
Fitting 113 distributions:  98%|█████████▊| 111/113 [01:19<00:01,  1.16it/s]
Fitting 113 distributions:  99%|█████████▉| 112/113 [01:20<00:01,  1.10s/it]
Fitting 113 distributions: 100%|██████████| 113/113 [01:21<00:00,  1.39it/s]

PR running time output:

Fitting 113 distributions:   0%|          | 0/113 [00:00<?, ?it/s]
Fitting 113 distributions:   1%|          | 1/113 [00:01<02:48,  1.51s/it]
Fitting 113 distributions:   4%|▍         | 5/113 [00:01<00:26,  4.07it/s]
...
Fitting 113 distributions:  93%|█████████▎| 105/113 [00:07<00:01,  5.51it/s]
Fitting 113 distributions:  96%|█████████▋| 109/113 [00:09<00:01,  3.71it/s]
Fitting 113 distributions:  98%|█████████▊| 111/113 [00:20<00:00,  3.71it/s]
Fitting 113 distributions:  99%|█████████▉| 112/113 [00:22<00:00,  1.02it/s]
Fitting 113 distributions: 100%|██████████| 113/113 [00:22<00:00,  4.97it/s]

[coveralls]

coverage: 92.727% (-7.3%) from 100.0%
when pulling b1418d8 on vitorandreazza:main
into 6a8c238 on cokelaer:main.

[cokelaer]

@vitorandreazza looks great to me. multiprocessing is always a bit tricky to test though. I will merge your work since it looks really good. All tests are good locally not in the continuous integration but i believe this is just not related. I'll update the version, readme and add your contribution straightway in the main branch. again thanks a lot.