Add wrapper to transform targets of (semi)supervised models

[ablaom]

This PR adds a model wrapper TransformedTargetModel implementing this suggestion of @CameronBieganek. Together with the Pipelinemodel already implemented (on the target branch) this wrapper renders the @pipeline macro redundant and whence resolves this Pluto notebook issue.

The doc-string for the new model appears below.

This PR is not breaking but I am basing it on the 0.19 release branch as it was convenient to use some utilities introduced there for the new pipelines.

To do:

• add deprecation warnings for @pipeline
• review test coverage

TransformedTargetModel(model; target=nothing, inverse=nothing, cache=true)

Wrap the supervised or semi-supervised model in a transformation of
the target variable.

Here target is either:

• The Unsupervised model that is to transform the training target.
  By default (inverse=nothing) the parameters learned by this
  transformer are also used to inverse-transform the predictions of
  model, which means target must implement the inverse_transform
  method. If this is not the case, specify inverse=identity to
  suppress inversion.

or

• A callable object for transforming the target, such as y -> log.(y). In this case
  a callable inverse, such as z -> exp.(z), should be specified.

Specify cache=false to prioritize memory over speed, or to guarantee data
anonymity.

Specify inverse=identity if model is a probabilistic predictor, as
inverse-transforming sample spaces is not supported. Alternatively,
replace model with a deterministic model, such as Pipeline(model, y -> mode.(y)).

Examples

A model that normalizes the target before applying ridge regression,
with predictions returned on the original scale:

@load RidgeRegressor pkg=MLJLinearModels
model = RidgeRegressor()
tmodel = TransformedTargetModel(model, target=Standardizer())

A model that instead applies a static log transformation to the data, again
returning predictions to the original scale:

tmodel2 = TransformedTargetModel(model, target=y->log.(y), inverse=z->exp.(y))

[ablaom]

@CameronBieganek Would be great if you could comment on the doc-string. If you also have time for a more detailed review over the next two weeks, let me know.

[codecov-commenter]

Codecov Report

Merging #678 (e49a339) into for-0-point-19-release (0f79fe9) will decrease coverage by 2.35%.
The diff coverage is 84.90%.

@@                    Coverage Diff                     @@
##           for-0-point-19-release     #678      +/-   ##
==========================================================
- Coverage                   85.84%   83.48%   -2.36%     
==========================================================
  Files                          40       41       +1     
  Lines                        3610     3173     -437     
==========================================================
- Hits                         3099     2649     -450     
- Misses                        511      524      +13     

Impacted Files	Coverage Δ
src/MLJBase.jl	100.00% <ø> (ø)
src/composition/models/transformed_target_model.jl	84.90% <84.90%> (ø)
src/sources.jl	70.00% <0.00%> (-18.00%)	⬇️
src/data/datasets.jl	86.84% <0.00%> (-13.16%)	⬇️
src/measures/continuous.jl	87.80% <0.00%> (-8.03%)	⬇️
src/show.jl	29.92% <0.00%> (-6.72%)	⬇️
src/measures/measures.jl	68.29% <0.00%> (-5.40%)	⬇️
src/measures/probabilistic.jl	58.46% <0.00%> (-4.70%)	⬇️
src/composition/models/inspection.jl	95.83% <0.00%> (-4.17%)	⬇️
src/measures/finite.jl	93.99% <0.00%> (-4.14%)	⬇️
... and 27 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 0f79fe9...e49a339. Read the comment docs.