BUG: mne.set_bipolar_reference function to account for bad channels

[moritz-gerster]

Reference issue

Fixes #11243.

What does this implement/fix?

When setting the bipolar reference between one good and one bad channel, the final bipolar channel is currently marked as good.

Bad channels are marked as bad because they contain artifacts. The bipolar reference between a bad channel and a good channel can never yield a good channel as a result. Therefore, the resulting bipolar channels should be marked as bad by mne.

Scenarios

Let us assume we have anode = ['Fc5.'] and cathode=['Fc3.'] and we want to create the bipolar channel ['Fc5.-Fc3.']. However, in our example ['Fc5.'] is marked as a bad channel. Currently, the resulting bipolar channel ['Fc5.-Fc3.'] is returned as a good channel. With our fix:

Scenarios

1. on_bad="warn" (default):
   set_bipolar_reference(raw, anode, cathode, ch_name=bipolar_chname, on_bad="warn")
   RuntimeWarning: Bipolar channels are based on bad channels: ['Fc5.-Fc3.'].

2. on_bad="raise":
   set_bipolar_reference(raw, anode, cathode, ch_name=bipolar_chname, on_bad="raise")
   ValueError: Bipolar channels are based on bad channels: ['Fc5.-Fc3.'].

3. on_bad="ignore":
   set_bipolar_reference(raw, anode, cathode, ch_name=bipolar_chname, on_bad="ignore")
   Added the following bipolar channels:
   Fc5.-Fc3.

Result

If on_bad is "raise" or "warn", the bipolar channel ['Fc5.-Fc3.'] will be marked as bad.

[larsoner]

Test failures look real and indicative of some problem

[moritz-gerster]

We tried everything to get as many tests pass as possible but some are still failing. @larsoner do you have some tips for us?

[larsoner]

Locally if I git merge upstream/main at least pytest mne/channels mne/io passes. Can you click the "allow edits by maintainers"? Then I can push this update and CIs should run again

[moritz-gerster]

Locally if I git merge upstream/main at least pytest mne/channels mne/io passes. Can you click the "allow edits by maintainers"? Then I can push this update and CIs should run again

@larsoner sure, but I think I checked this box already?