BUG: Fix dev_head_t for EEG data

[larsoner]

Seems like a bug that for EEG-only data formats like egi-mff that info["dev_head_t"] is not None. There is no MEG device so device to head transform. We currently set it to the identity matrix.

This is also relevant for example when concatenating MEG and EEG data -- if the MEG data has a dev_head_t (and it should) and the EEG data has dev_head_t of None everything works okay. If the EEG data has the identity matrix we get:

...
    meg.add_channels([eeg])
  File "/home/larsoner/python/mne-python/mne/channels/channels.py", line 741, in add_channels
    new_info = _merge_info(infos, force_update_to_first=force_update_info)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<decorator-gen-34>", line 12, in _merge_info
  File "/home/larsoner/python/mne-python/mne/_fiff/meas_info.py", line 3089, in _merge_info
    raise ValueError(msg)
ValueError: Measurement infos provide mutually inconsistent dev_head_t

Opening as draft because I think we need to go through and find all EEG-only formats and fix this, so far I've only done it for EGI-MFF.

• Ensure CIs happy for MFF
• Add to other EEG formats with tests

[sappelhoff]

Could it be that this is also the reason that the mne-bids test pipeline is currently failing?

https://github.com/mne-tools/mne-bids/actions/runs/13351688224/job/37289052292

FAILED mne_bids/tests/test_read.py::test_get_head_mri_trans - ValueError: Coordinate frame not supported: None
================== 1 failed, 288 passed in 264.35s (0:04:24) ===================