add "duration" field to audio cells

As it was done with width/height for image cells: https://github.com/huggingface/dataset-viewer/pull/600

It will help to show additional information in the dataset viewer, in particular: highlighting a row will select the appropriate bar in the durations histogram.

Start by tracing how width and height are added for image cells, using the referenced pull request 600 as the entry point, then find the corresponding audio-cell representation. Done means audio cells expose a duration value that the dataset viewer can use to select the matching bar in the durations histogram.

Written by the indexing model from the issue text.