Parallel evaluation of tasks

[Neeratyoy]

What does this PR implement/fix? Explain your changes.

Uses joblib to evaluate different folds of a task in parallel, in order to speed-up the purely sequential evaluations of tasks right now.

How should this PR be tested?

The unit tests added to test the accuracy of the distribution of evaluation and collation of results.

[codecov-io]

Codecov Report

Merging #1020 (c2cbfe2) into develop (ab793a6) will increase coverage by 0.13%.
The diff coverage is 84.48%.

@@ Coverage Diff @@## develop #1020 +/- ##
===========================================
+ Coverage 88.04% 88.18% +0.13% 
===========================================
Files 36 36 Lines 4568 4594 +26 ===========================================
+ Hits 4022 4051 +29 + Misses 546 543 -3 

Impacted Files	Coverage Δ
openml/runs/functions.py	83.61% <72.72%> (+0.15%)	⬆️
openml/config.py	84.78% <100.00%> (+5.10%)	⬆️
openml/testing.py	84.24% <100.00%> (+0.70%)	⬆️
openml/utils.py	90.66% <0.00%> (-0.67%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ab793a6...17e3916. Read the comment docs.