Add STLDecomposer to multiseries pipelines

[remyogasawara]

Resolves #4298

Acceptance Criteria (AC)

• Pipelines generated during AutoMLSearch via make_pipeline for multiseries time series regression include one pipeline with the STLDecomposer and one pipeline without, following the with/without pattern we use for current time series regression problems

[codecov]

Codecov Report

Patch coverage: 100.0% and project coverage change: +0.1% 🎉

Comparison is base (1329988) 99.7% compared to head (28e2cdb) 99.7%.

Additional details and impacted files

@@           Coverage Diff           @@
##            main   #4299     +/-   ##
=======================================
+ Coverage   99.7%   99.7%   +0.1%     
=======================================
  Files        357     357             
  Lines      39577   39587     +10     
=======================================
+ Hits       39457   39467     +10     
  Misses       120     120             

Files Changed	Coverage Δ
evalml/pipelines/component_graph.py	99.8% <ø> (ø)
...nents/transformers/preprocessing/stl_decomposer.py	100.0% <100.0%> (ø)
evalml/pipelines/time_series_pipeline_base.py	100.0% <100.0%> (ø)
evalml/pipelines/utils.py	99.7% <100.0%> (+0.1%)	⬆️
...valml/tests/automl_tests/test_default_algorithm.py	100.0% <100.0%> (ø)
...lml/tests/automl_tests/test_iterative_algorithm.py	100.0% <100.0%> (ø)
evalml/tests/pipeline_tests/test_pipeline_utils.py	99.6% <100.0%> (ø)

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

[eccabay]

This is hard to follow 😅 can you add an explanatory comment?

[eccabay]

Out of curiosity, why is this necessary? What was the situation where the columns weren't the same?

[remyogasawara]

The predictions weren't getting the corresponding series ID values as the column names and that's needed since the decomposer uses this to select the correct value. Before this was causing the decomposer to return NaN values. @christopherbunn figured that out so he might have more info.

[christopherbunn]

The predictions that are generated do not have the series ID values as their column names. Copying these names over is required so we can inverse_transform from the decomposer.

[eccabay]

The way determine_periodicity is set up, we're currently detecting a "period" on the single stacked target data column. I'm worried that that's too brittle, it could cause weird issues in the future. Could you put this in a conditional branch to ensure we only run it in the single series case, and for now just always add the decomposer for multiseries? We'll have to come back and revisit, but that should be ok for the MVP.