Multi-annotation evaluation, dataframe results

[bmcfee]

Two issues, bundled together here because they make sense:

1. It would be cool if jams.eval could automatically convert the results dict into a dataframes.
2. Also, jams.eval could support comparisons of multiple estimates to a reference.

These two can play nicely together if the index of the dataframe is drawn from estimate annotation. This way, you could do something like

results_df = jams.eval.chord(ref_ann, est_ann1, est_ann2, est_ann3, dataframe=True)

And get a nice table of results to make comparison easy.