can't pickle Epochs with preload=False

To replicate:

from pathlib import Path
import mne
import pickle

sample_dir = Path(mne.datasets.sample.data_path())
sample_fname = sample_dir / 'MEG' / 'sample' / 'sample_audvis_raw.fif'

raw = mne.io.read_raw_fif(sample_fname)
raw.crop(tmax=60)

events = mne.find_events(raw, stim_channel='STI 014')
event_id = {'auditory/left': 1}
epochs = mne.Epochs(raw, events=events, event_id=event_id,
                    tmin=-0.2, tmax=0.5, baseline=(None, 0),
                    preload=True)

epochs.save('toto-epo.fif', overwrite=True)
epochs = mne.read_epochs('toto-epo.fif', preload=False)

# del epochs._raw[0].fid  # uncomment this line to fix it

fname = "toto.pkl"
with open(fname, 'wb') as fid:
    pickle.dump(epochs, fid)

it's due to the fid in epochs._raw[0].fid

@larsoner thoughts?

reported by @gemeinl in braindecode

Start by running the provided Python reproduction with MNE's sample dataset and inspect the loaded Epochs object's epochs._raw[0].fid after read_epochs(..., preload=False). Done means an Epochs object loaded with preload=False can be serialized with pickle without requiring the manual fid deletion shown in the report.