Refactor OPM/Neuromag grouping

[PragnyaKhandelwal]

Reference issue (if any)

Follow‑up to #13866.

What does this implement/fix?

• Makes tangential OPM topomaps use unsigned RMS magnitude (aligned with Neuromag grads).
• Uses per‑group vlim and non‑diverging colormap for tangential maps.
• Adds a shared helper to unify grouping/merging logic for OPM and Neuromag‑style channels.
• Updates OPM grouped topomap tests for new axes/colorbar behavior.

Additional information

• Draft PR; please review after ENH: group triaxial OPM topomaps by orientation #13866 merges (I will rebase then).

[larsoner]

Failures look related. The ubuntu-latest / pip job runs the most complete set of tests, so you can look here for a good list of failures:

https://github.com/mne-tools/mne-python/actions/runs/26439596935/job/77830628215?pr=13921

Once you have these working locally feel free to push fixes!

[larsoner]

@PragnyaKhandelwal CircleCI only builds modified examples. Can you make some minor changes (can just be trivial whitespace / word-wrapping) to some examples you'd expect to change due to these changes? From a quick look I think that would be at least something that uses evoked.plot with triaxial OPM data (kernel one?)

[PragnyaKhandelwal]

Hey @larsoner! Thanks for the heads-up on how CircleCI handles the examples.
To make the review process a bit easier, I actually just folded this entire refactor branch directly into PR #13866! Since kernel_phantom.py is already modified in that branch, CircleCI successfully picked it up and generated the updated visual artifacts over there.
I'm going to close this one out so we can keep all the discussion and review momentum centralized in #13866. See you over on that thread!