Adding SimpleImputer, IterativeImputer and KNNImputer to the config space.

[gketronDS]

What does this PR do?

Adds sklearn.impute functions into TPOT2's configuration search space.

Where should the reviewer start?

See changes below to get_configspace.py and imputers.py, there are not many.

How should this PR be tested?

Simple, Iterative, and KNN Imputers passed pytest for the config folder. I wasn't able to install scikit-learn-intelex to confirm it passes everything in Tpot2.

Any background context you want to provide?

Allows us to soft-code preprocessing, and allow for additional preprocessing optimization within TPOT.

What are the relevant issues?

N/A

Questions:

• Do the docs need to be updated?
  May need to be added to the next version update.
• Does this PR add new (Python) dependencies?
  No, just uses scikit-learn.

[perib]

some estimators with the iterative imputation search space return fail when sample_posterior=True. The estimator must support "return_std" in the predict function for if sample_posterior=True.

I would recommend either add a conditional to only search sample_posterior=True if BayesianRidge is selected, or just keep sample_posterior to the default value of False.

[gketronDS]

Comments from Pedro: "Putting the conditional in the parser is tricky because it means that posterior=True or False is equivalent when using the other three methods, but TPOT may not know this and evaluate it twice. instead, I would recommend putting in the config_space function similar to how it was done in the logistic regression example."

Fixed in [Conditional Sample Posterior Added for Iterative Imputer]