[ENH] extending PR #2017 to multiclass case

[IRKnyazev]

Reference Issues/PRs
See also #2016

What does this implement/fix? Explain your changes.
This is extending the previous PR which missed out the multiclass case for linear classifiers, here an distance is made inversely correlated to distance and for RDST the other two features are kept positively correlated.

Any other comments?
@baraline noticed that I missed the multiclass correction and explained that coefs are one vs all for this usecase so just need to grab the class coefs and manipulate them appropriately.
This is a fresh PR resolving merge conflicts which were present in PR 2018.

[aeon-actions-bot]

Thank you for contributing to aeon

I have added the following labels to this PR based on the title: [ $\mathsf{\text{enhancement}}$ ].
I have added the following labels to this PR based on the changes made: [ $\mathsf{\text{visualisation}}$ ]. Feel free to change these if they do not properly represent the PR.

The Checks tab will show the status of our automated tests. You can click on individual test runs in the tab or "Details" in the panel below to see more information if there is a failure.

If our pre-commit code quality check fails, any trivial fixes will automatically be pushed to your PR unless it is a draft.

Don't hesitate to ask questions on the aeon Slack channel if you have any.

PR CI actions

These checkboxes will add labels to enable/disable CI functionality for this PR. This may not take effect immediately, and a new commit may be required to run the new configuration.

• Run pre-commit checks for all files
• Run all pytest tests and configurations
• Run all notebook example tests
• Run numba-disabled codecov tests
• Stop automatic pre-commit fixes (always disabled for drafts)
• Push an empty commit to re-run CI checks

[baraline]

Quick question and we should be good to go !

[baraline]

Thanks, and indeed the linear case is a bit weird to make out for RDST, as we're trying to explain what would likely be a non-linear relationship with linear coefs (without any kind of kernel trick) ... This is likely why on some datasets, using non-linear classifiers with RDST gives way more accuracy than the linear one (and vice versa)