FIX, MAINT: Implement 'everything follows X' and namespace checks for KNN

[david-cortes-intel]

Description

This PR:

• Implements the logic of 'everything follows X' for KNN classes.
• Implements the logic where class predictions from classifiers follow the 'y' namespace (e.g. so that it can return strings when fitting on GPU).
• Adds array API namespace and device checks for KNN methods that come after .fit() in order to throw informative Python exceptions instead of segfaults, the same way scikit-learn would do.
• Fixes some issues with internal functions that were not working with array_api_strict.

Includes the changes from #3117 since they are also necessary for KNN classes.

---

Checklist:

Completeness and readability

• I have commented my code, particularly in hard-to-understand areas.
• Git commit message contains an appropriate signed-off-by string (see CONTRIBUTING.md for details).
• I have resolved any merge conflicts that might occur with the base branch.

Testing

• I have run it locally and tested the changes extensively.
• All CI jobs are green or I have provided justification why they aren't.
• I have extended testing suite if new functionality was introduced in this PR.

[david-cortes-intel]

/intelci: run

[david-cortes-intel]

/azp run Nightly

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[david-cortes-intel]

CI failures in BasicStatistics are unrelated to the changes here and should be solved with this PR:
#3128

[david-cortes-intel]

/intelci: run

[david-cortes-intel]

/azp run Nightly

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[codecov]

Codecov Report

❌ Patch coverage is 28.94737% with 54 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
sklearnex/neighbors/common.py	42.85%	12 Missing and 4 partials ⚠️
sklearnex/neighbors/knn_classification.py	23.80%	9 Missing and 7 partials ⚠️
sklearnex/neighbors/knn_regression.py	17.64%	8 Missing and 6 partials ⚠️
sklearnex/neighbors/knn_unsupervised.py	20.00%	5 Missing and 3 partials ⚠️

Flag	Coverage Δ
azure	77.50% <28.94%> (-1.55%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
sklearnex/neighbors/_lof.py	98.01% <ø> (-1.99%)	⬇️
sklearnex/neighbors/knn_unsupervised.py	86.17% <20.00%> (-8.02%)	⬇️
sklearnex/neighbors/knn_regression.py	79.36% <17.64%> (-9.93%)	⬇️
sklearnex/neighbors/common.py	85.83% <42.85%> (-4.53%)	⬇️
sklearnex/neighbors/knn_classification.py	84.76% <23.80%> (-10.05%)	⬇️

... and 37 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[david-cortes-intel]

/azp run Nightly

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[david-cortes-intel]

/intelci: run

[avolkov-intel]

What is the logic of this if-statement? As I understand it previously it was that neigh_dist and neigh_ind originally have numpy type and we only need to convert them if y is not a numpy. Is it correct? If yes do we need the same logic after the array-api update?

[david-cortes-intel]

This is how it was before if you look at the changes. I guess the purpose is to have them work with the other arrays.

[avolkov-intel]

I think we also need some explanation about why do we need this numpy check

[david-cortes-intel]

It has a different codepath for numpy with operations that are not supported by array API.

[avolkov-intel]

Left a few comments mostly related to clarification about the current logic. Also PR needs to be rebased to fix some CI failures