This set of tests were written to use the `unittest` framework
from the standard library but we invoke them through pytest,
which provides better reporting.

Remove the unittest dependency here to align with changes made
to other tests.

As a minimal change, keep the TestParser class, but remove the
inheritance from `unittest.TestCase`. The pytest package instead
finds tests based on general introspection. By marking the setUp
method, which has a special meaning in the unittest framework as
a pytest fixture with `autouse=True` we ensure it is called before
every test to set up the member variable.

Move the lexer fixture into the `try_parse` helper. None of the
tests need subsequent access to the lexer state, so there's no
reason for it to be shared.

Without shared state, the TestParser class can itself be removed.
It was necessary under the `unittest` framework, but `pytest` can
use module functions instead. This simplifies the code and saves
an indent level.

The parser creates `Score` objects from atf records so marked.
Verify this in the related tests.

These tests include atf snippets for features which the lexer
accepts, but the parser currently skips. They were marked with
a commented-out @Skip decorator. With the migration to pytest
these should become @pytest.skip instead.

Instead I moved the annotation to a TODO comment, mostly because
I like the style better. I also added basic asserts on the parser
result to silence warnings.

For an example of the '={' atf syntax, see the sample corpus, and
e.g. http://oracc.iaas.upenn.edu/saao/saa10/P334031/html

Check the attributes on the Score object returned from the
parser.

Parametrize the test with the different configurations described
in http://oracc.museum.upenn.edu/doc/help/editinginatf/scores/

This verifies more of the documented variants are accepted.

Restore the skip annotation for tests referencing features
supported by the lexer but not the parser. While there's
some value in verifying the parser accepts the input strings,
it's more useful to mark the state of the actual implementation
in the pytest output.