ENH Speed up sgdclassifier example plot_sgd_early_stopping.py

[johgreen]

Reference Issues/PRs

Speeding up ../examples/linear_model/plot_sgd_early_stopping.py in #21598

What does this implement/fix? Explain your changes.

Reduced n_samples from 10000 to 5000 to speed up runtime.

[ogrisel]

Can you please fix the code style to get the linting work in the CI status summary? Look at the logs of the failed CI jobs for details.

Please also report your analysis of the before/after of your proposed change:

• observed runtime speed-up;
• preservation of the results and conclusion of the example.

[adrinjalali]

Interestingly, at least on the CI, this example doesn't run faster than what we have.

You also need to run black on your code as I said before @johgreen

[ogrisel]

I think this example might be completely reworked to be much more elegant and efficient once me make some progress on the callback API #16925 that itself is being worked on by @jeremiedbb.

[adrinjalali]

I agree @ogrisel , but this cuts over 30s from the example, we can merge as is, and let @jeremiedbb rework the example as a part of the callback work. WDYT?

[adrinjalali]

ping @ogrisel

[adrinjalali]

As is, this PR doesn't seem to change the speed of the example that much (33s vs 30s)

[glemaitre]

Closing this PR since we have already tackled the problem (23s in main).