feat(core): implement a more robust set_params()

[deepyaman]

Related to #135 and #136

Before this PR, it wasn't possible to call set_params() on Step objects, and Recipe.set_params() only trivially set the steps param. With this PR, you can actually set the parameters listed by Recipe.get_params() (and Step.get_params()), which is more robust and akin to scikit-learn.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 96.66667% with 2 lines in your changes missing coverage. Please review.

Project coverage is 88.70%. Comparing base (cfd6dc1) to head (effadce).

Files with missing lines	Patch %	Lines
ibis_ml/core.py	94.28%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #162      +/-   ##
==========================================
+ Coverage   87.37%   88.70%   +1.32%     
==========================================
  Files          27       27              
  Lines        1964     2009      +45     
==========================================
+ Hits         1716     1782      +66     
+ Misses        248      227      -21     

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

[jcrist]

Can you add a test that constructs a sklearn pipeline with a recipe, then calls the pipeline's get_params and set_params as sklearn would to ensure everything is plumbed through correctly?

[deepyaman]

I didn't add anything here since test_can_use_in_sklearn_pipeline already covers that case (and was failing until got it "right").

[jcrist]

That's probably the better place for it, but that test doesn't hit the setting nested parameters workflow directly. Can you add one that does something like:

pipe = ...
pipe.set_params(recipe__step__something="foo")
# then assert the nested thing was set

[deepyaman]

It is hitting it, but I've made more explicit in effadce

[deepyaman]

@jcrist if this looks good, I will go ahead and merge!