FIX Fixes visualization for nested meta-estimators

[thomasjpfan]

Reference Issues/PRs

Fixes #21267

What does this implement/fix? Explain your changes.

This PR changes the implementation a little to only work on estimators. This fixes the original issue:

from sklearn.gaussian_process.kernels import ExpSineSquared
from sklearn.kernel_ridge import KernelRidge
from sklearn.ensemble import VotingClassifier
from sklearn.model_selection import RandomizedSearchCV

from sklearn import set_config

set_config(display="diagram")
kernel_ridge = KernelRidge(kernel=ExpSineSquared())

kernel_ridge_tuned = RandomizedSearchCV(
    kernel_ridge,
    param_distributions={},
)
kernel_ridge_tuned

this PR

main

Any other comments?

There was also issues with nesting meta-estimators that is resolved:

from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import RandomizedSearchCV
from sklearn.ensemble import VotingClassifier
from sklearn.ensemble import RandomForestClassifier

vc = VotingClassifier(
    [("log_reg", LogisticRegression()),
    ("rf", RandomForestClassifier())]
)

vc_tuned = RandomizedSearchCV(
    vc,
    param_distributions={},
)
vc_tuned

this PR

main

[jeremiedbb]

Shouldn't get_params be called with deep=False ?
I guess it's what "__" not in key is suppose to mimic

[thomasjpfan]

I agree setting deep=False works here. I updated the PR with your suggestion.