Inside `SimpleSeq.from_file`, when someone specifies a
`column_map` argument, we only do the mapping if a column
name in the loaded csv file is a key in `column_map`.
Otherwise we leave the column as-is, and if it's not one of
the columns we need for the annotations
("onset_s", "offset_s", "label"),
it gets thrown away.

Adds `default_label` argument to `from_file` method,
a string that defaults to `'-'`
and if the `pandas.DataFrame`
loaded from the csv file does not already have a
'label' column, adds that column with the
`default_label` value as the label in every row.

This makes it possible to load segmentations saved
in csv files that do not have labels,
just start and stop times of the segments.

As described in issue, and also:
Add example data files to dir
Remove crowsetta.data
Rewrite tests/test_data.py as tests/test_example.py

…'t have expected names.

…xample

…pfile

…nd get a Sequence of length 0

…les.show and crowsetta.example