FIX: incremental estimators tests (backport #1998)

[mergify]

Description

fixes incremental estimators tests.
Output array should be converted to numpy array before setting tol dependent on the dtype, otherwise the tol settings will not be set. Found on #1861

---

Checklist to comply with before moving PR from draft:

PR completeness and readability

• I have reviewed my changes thoroughly before submitting this pull request.
• I have commented my code, particularly in hard-to-understand areas.
• I have updated the documentation to reflect the changes or created a separate PR with update and provided its number in the description, if necessary.
• Git commit message contains an appropriate signed-off-by string (see CONTRIBUTING.md for details).
• I have added a respective label(s) to PR if I have a permission for that.
• I have resolved any merge conflicts that might occur with the base branch.

Testing

• The unit tests pass successfully.
• I have run it locally and tested the changes extensively.

Performance

• I have measured performance for affected algorithms using scikit-learn_bench and provided at least summary table with measured data, if performance change is expected.
• I have provided justification why performance has changed or why changes are not expected.

---

This is an automatic backport of pull request #1998 done by [Mergify](https://mergify.com).

[samir-nasibli]

/intelci: run

[samir-nasibli]

CI fail are not related with the changes in the PR