mne.grand_average for source estimates

Describe the new feature or enhancement

For consistency between sensor and source analysis, it would be nice to allow for mne.grand_average as input data a list of mne.SourceEstimate. Users might be fooled into believing that grand avering source estimates it not suggested.

Describe your proposed implementation

the all_inst parameter that takes the evoked datasets of mne.grand_average should allow for list of mne.SourceEstimate. All source estimates in the list (https://mne.tools/stable/generated/mne.SourceEstimate.html) should have the same number of vertices. The nave parameter of the evokeds could be set to 1 as a workaround. interpolate_bads and drop_bads should check if evokeds equal Source Estimate and if True, both parameters should be set to False.

Describe possible alternatives

An alternative could be to implement a separate function for the source estimate class [mne.SourceEstimate], that computes grand_averages on a list of source estimates.

Additional context

No response

Start with the mne.grand_average entry point and the SourceEstimate API described in the issue, then inspect how the current function handles Evoked inputs and the all_inst, interpolate_bads, and drop_bads parameters. Done means a list of SourceEstimate objects with matching vertex counts is accepted, with the stated bad-channel options handled appropriately, and coverage is added for the new behavior.