Add auto base and auto abstract

[GooseIt]

Before submitting (must do checklist)

• Did you read the contribution guide?
• Did you update the docs? We use Numpy format for all the methods and classes.
• Did you write any new necessary tests?
• Did you update the CHANGELOG?

Proposed Changes

See issue#1023

Closing issues

closes #1023

[codecov-commenter]

Codecov Report

Merging #1114 (37f2106) into master (cdf1ee3) will increase coverage by 18.66%.
The diff coverage is 85.71%.

❗ Current head 37f2106 differs from pull request most recent head 881f38e. Consider uploading reports for the commit 881f38e to get more accurate results

📣 This organization is not using Codecov’s GitHub App Integration. We recommend you install it so Codecov can continue to function properly for your repositories. Learn more

@@             Coverage Diff             @@
##           master    #1114       +/-   ##
===========================================
+ Coverage   68.39%   87.06%   +18.66%     
===========================================
  Files         170      170               
  Lines        9335     9347       +12     
===========================================
+ Hits         6385     8138     +1753     
+ Misses       2950     1209     -1741     

Impacted Files	Coverage Δ
etna/auto/auto.py	90.74% <85.71%> (+48.21%)	⬆️
etna/models/nn/rnn.py	100.00% <0.00%> (ø)
etna/models/sarimax.py	95.00% <0.00%> (+0.71%)	⬆️
etna/models/prophet.py	99.13% <0.00%> (+0.86%)	⬆️
etna/models/holt_winters.py	98.85% <0.00%> (+1.14%)	⬆️
etna/models/catboost.py	100.00% <0.00%> (+1.21%)	⬆️
etna/transforms/math/add_constant.py	97.95% <0.00%> (+2.04%)	⬆️
etna/analysis/outliers/density_outliers.py	97.95% <0.00%> (+2.04%)	⬆️
etna/models/tbats.py	95.00% <0.00%> (+3.33%)	⬆️
etna/metrics/base.py	93.67% <0.00%> (+3.79%)	⬆️
... and 61 more

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[Mr-Geekman]

Look at comments above.

[Mr-Geekman]

Look at comments above.

[GooseIt]

I've moved most of __init__ from Auto to AutoBase on the basis that every field initialized there except pool will be shared between Auto and Tune classes with same functionality.

[Mr-Geekman]

What about refactoring test_auto?

[GooseIt]

What about refactoring test_auto?

As classes AutoBase and AutoAbstract both have unimplemented abstractmethods (namely fit), I see no tests that can be written to evaluate their performance (these classes cannot even be instantiated).
There is no need to write any additional tests for Auto neither, as there is no new functional added.

If I am wrong, what the refactored tests should check?

[Mr-Geekman]

We have methods summary and top_k in AutoBase. We should check them. In test_auto we have tests test_summary, test_summary, we should change them to work with AutoBase.

You doesn't have to instantiate classes, tests are written using mocks.