reduce memory_footprint for sparse PCA transform

[Intron7]

The sparse PCA still densified X during the transform step. This defeats the purpose of a sparse PCA in a sense. However

precomputed_mean_impact = self.mean_ @ self.components_.T
mean_impact = cp.ones((X.shape[0], 1)) @ precomputed_mean_impact.reshape(1, -1)
X_transformed = X.dot(self.components_.T) -mean_impact

is the same as

X = X - self.mean_
X_transformed = X.dot(self.components_.T)

The new implementation is faster (but mainly due to the fact that we don't have to rely on cupy's to_array()) and uses a lot less memory.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[lowener]

/ok to test

[Intron7]

ok I added the right year in the cupyright

[lowener]

/ok to test

[Intron7]

I think I can't fix the issue on the PR since it's in regards to the new layout

[dantegd]

@Intron7 yeah, I think the issue is that we just merged the PR to nest the python package one level lower #5944 so it should be an easy resolution of conflict I hope

[Intron7]

@lowener I merged in the updated main branch

[dantegd]

/ok to test

[dantegd]

/ok to test

[dantegd]

/ok to test

[dantegd]

/merge