Update pytest workflow and docker dist environements from Ubunut 18.04 to 20.04

[eddiebergman]

The docker distribution was failing due to Python 3.6 being the default Python distribution on ubuntu 18.04 as seen in issue #1193. The tests were also running on ubuntu 18.04 which should be upgraded the latest LTS version by ubuntu.

Updated the docker file to use ubuntu:20.04 instead.

Edit:
Upgrading the pytest workflow from 18.04 to 20.04 causes numerical differences in 4 tests where it seems all models perform better than the previous hardcoded test results. (log)

FAILED test/test_pipeline/components/regression/test_sgd.py::SGDComponentTest::test_default_boston
FAILED test/test_pipeline/components/regression/test_sgd.py::SGDComponentTest::test_default_boston_iterative_fit
FAILED test/test_pipeline/components/regression/test_sgd.py::SGDComponentTest::test_default_boston_iterative_sparse_fit
FAILED test/test_pipeline/components/regression/test_sgd.py::SGDComponentTest::test_default_boston_sparse

[codecov]

Codecov Report

Merging #1198 (8700e2b) into development (96b9ad0) will increase coverage by 0.48%.
The diff coverage is n/a.

@@               Coverage Diff               @@
##           development    #1198      +/-   ##
===============================================
+ Coverage        88.15%   88.64%   +0.48%     
===============================================
  Files              138      142       +4     
  Lines            10866    12145    +1279     
===============================================
+ Hits              9579    10766    +1187     
- Misses            1287     1379      +92     

Impacted Files	Coverage Δ
autosklearn/estimators.py	92.89% <0.00%> (-0.47%)	⬇️
...ta_preprocessing/rescaling/quantile_transformer.py	100.00% <0.00%> (ø)
...line/components/data_preprocessing/feature_type.py	88.11% <0.00%> (ø)
...nents/data_preprocessing/feature_type_numerical.py	90.32% <0.00%> (ø)
...pipeline/components/data_preprocessing/__init__.py	82.97% <0.00%> (ø)
...nts/data_preprocessing/feature_type_categorical.py	90.90% <0.00%> (ø)
autosklearn/experimental/askl2.py	85.34% <0.00%> (+0.12%)	⬆️
autosklearn/evaluation/abstract_evaluator.py	93.07% <0.00%> (+0.15%)	⬆️
autosklearn/pipeline/regression.py	95.00% <0.00%> (+0.31%)	⬆️
...ture_preprocessing/select_percentile_regression.py	97.43% <0.00%> (+1.43%)	⬆️
... and 9 more

---

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 96b9ad0...8700e2b. Read the comment docs.