Reuse the `pre-commit` linter config in the CI runners

At the moment, our CI runs linting jobs using a custom script build_tools/lintin.sh with manually pip installed dependencies in the Circle CI and Azure Pipelines CI configuration files.

Those linter tool versions are not always consistent with what we configure in .pre-commit-config.yaml and that can cause problems such as the following: https://github.com/scikit-learn/scikit-learn/pull/32499#issuecomment-3402244557.

I think we should refactor our CI config to reuse the pre-commit command, for instance by running:

pre-commit run --all-files

Instead of our custom build_tools/linting.sh script. However, this might not be that easy:

• there are probably linting failures to fix or config to update to ignore specific files or folders;
• there might be CI integration to update, in particular the .github/workflows/bot-lint-comment.yml and the ./build_tools/get_comment.py script that generate a helpful user facing comments will need to be reworked and adapted to work with the output of the pre-commit command.

Start by comparing the linting steps in the Circle CI and Azure Pipelines configuration files with .pre-commit-config.yaml and build_tools/linting.sh. Then inspect .github/workflows/bot-lint-comment.yml and build_tools/get_comment.py to understand how lint output is reported. Done means CI reuses the pre-commit configuration consistently, required lint failures or exclusions are addressed, and user-facing comments still work.