Add geom_wave

[bvreede]

If audio data is present, geom_wave would plot a spectrogram over the utterances.

[mdingemanse]

Important to note that we cannot assume geom_wave can (always) be mapped to participant. Since audio data is frequently stereo (=two channels) but not channel-separated (=one speaker per channel), it will not always be possible to position geom_wave over the turn. So it will sometimes appear in a layer that shares the time scale but not the participant aesthetic.

Also see #55, which provides another reason to think of geom_wave as not necessarily inheriting the participant aesthetic.

[mdingemanse]

Note that audio data might be actual .wav files but as @liesenf points out, more likely is plottable data in the source csv file, e.g. as generated by scikit-talk. Related issue here elpaco-escience/scikit-talk#23 (thanks @bvreede)