Feat/global naive models

[dennisbader]

Summary

• Adds global naive models that can predict on single, multiple, uni- and or multivariate series.
  • GlobalNaiveAggregate: Accepts an agg_fn for aggregating the last input_chunk_length target values. By default it uses the mean. Depending on the input_chunk_length, and output_chunk_length, this model can be identical to NaiveMean, NaiveMovingAverage.
  • GlobalNaiveDrift: Depending on the input_chunk_length, and output_chunk_length, this model can be identical to NaiveDrift. It also support moving/autoregressive drift
  • GlobalNaiveSeasonal: Depending on the input_chunk_length, and output_chunk_length, this model can be identical to NaiveSeasonal. It also supports repeating value mode (output_chunk_length > n (forecast horizon))

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 92.85714% with 12 lines in your changes are missing coverage. Please review.

Project coverage is 93.97%. Comparing base (4de71e4) to head (5d1624e).

Files	Patch %	Lines
darts/models/forecasting/global_baseline_models.py	92.37%	9 Missing ⚠️
...arts/models/forecasting/torch_forecasting_model.py	94.28%	2 Missing ⚠️
darts/utils/historical_forecasts/utils.py	50.00%	1 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2261      +/-   ##
==========================================
- Coverage   93.98%   93.97%   -0.02%     
==========================================
  Files         135      136       +1     
  Lines       13525    13647     +122     
==========================================
+ Hits        12712    12825     +113     
- Misses        813      822       +9     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[madtoinou]

Nice! This will make it much easier for user to implement very simple models as well.

Some small changes and it should be ready for merge.

[madtoinou]

LGTM, thanks for this great PR @dennisbader 🚀