Remove `split_tracks_df_by_lineage` and replace with it with `split_trees`

[JoOkuma]

I noticed this is a duplicated function, and split_trees is much faster; we could rename it to make it clearer.