Bug: event_repeated metadata incompatibility

mne.Epochs event_repeated parameter does not play well metadata:

import mne
import numpy as np
import pandas as pd
raw_data = np.random.randn(2, 10000)
info = mne.create_info(2, 1000.)
raw = mne.io.RawArray(raw_data, info)

opts = dict(raw=raw, tmin=0, tmax=.001, baseline=None)

No problem when there are no repeated events and metadata

events = [[0, 0, 1], [1, 0, 1]]
metadata = pd.DataFrame(events, columns=['onset', 'duration', 'value'])

epochs = mne.Epochs(events=events, metadata=metadata, **opts)

No problem when repeated events and no metadata

events = [[1, 0, 1], [1, 0, 1]]
epochs = mne.Epochs(events=events, event_repeated='drop', **opts)

problem when repeated events and metadata

events = [[1, 0, 1], [1, 0, 1]]
epochs = mne.Epochs(events=events, event_repeated='drop', metadata=metadata, **opts)

Start by running the provided reproducer with mne.Epochs, event_repeated='drop', and metadata. Trace how repeated events are removed and how metadata rows are aligned; done means the repeated-event case works with metadata without an incompatibility.