New feature: EEG standardization function

[christian-oreilly]

I'd like to add a function that allows to standardize the EEG so that it can be easily used to combine datasets with different sampling and spatial frequencies. Uniformizing the sampling frequency can be done easily with the resampling functions from MNE (https://mne.tools/stable/auto_tutorials/preprocessing/30_filtering_resampling.html).

For the spatial resampling, we need to interpolate on a standard EEG montage. The idea here is to

1. Get original montage (montage1)
2. Get montage to interpolate to (montage2)
3. Co-register montage2 with montage1 on a same head surface if these are not already co-registered (not in the code above)
4. Create a Raw structure containing with flat signals (zeros) for the data associated with montage2
5. Merge the two montages
6. Merge the two raw objects and set the raw montage
7. Flag all the channels from montage2 as 'bad' in the merged raw
8. Interpolate bad channels
9. Drop channels from montage 1

This is a snippet of code taken from one of my previous analyses that do more or less that:

    montage_egi = mne.channels.read_dig_fif(str(Path(subjects_dir) / template / "montages" / montage_name_egi))
    montage_egi.ch_names = ["E" + str(int(ch_name[3:])) for ch_name in montage_egi.ch_names]
    montage_egi.ch_names[128] = "Cz"

    raw_egi = preprocessed_raw(path, line_freqs[dataset], montage_egi,
                                rename_channel=False, apply_ica=apply_ica)
    events = process_events_resting_state(raw_egi, dataset, age)
    if events is None:
        return None

    montage_1010 = mne.channels.read_dig_fif(str(Path(subjects_dir) / template / "montages" / montage_name_1010))
    montage_1010.ch_names = ["S" + str(int(ch_name[3:])) for ch_name in montage_1010.ch_names]
    info_1010 = mne.create_info(montage_1010.ch_names, sfreq=500, ch_types="eeg")
    raw_1010 = mne.io.RawArray(np.zeros((len(montage_1010.ch_names), raw_egi.get_data().shape[1])),
                               info_1010, copy=None, verbose=False).set_montage(montage_1010)

    merged_montage = montage_egi + montage_1010
    merged_data = np.vstack([raw_egi.get_data(), raw_1010.get_data()])
    merged_info = mne.create_info(merged_montage.ch_names, sfreq=raw_egi.info["sfreq"], ch_types="eeg")
    merged_info['bads'] = montage_1010.ch_names
    merged_raw = mne.io.RawArray(merged_data, merged_info).set_montage(merged_montage)
    merged_raw = merged_raw.interpolate_bads()

    merged_raw.pick(montage_1010.ch_names)
    merged_raw.set_annotations(raw_egi.annotations)

(not complete and executable in its current form)

As an added bonus, adding this feature will put us in a very good place to close #131

[christian-oreilly]

As pointed out by @scott-huberty, we should consider using the feature implemented in mne-tools/mne-python#13044 whenever this PR merges.