Remove biggus test skipper: fix pp field data

[lbdreyer]

This PR removes a single skip_biggus test skipper.

Work done:

• In the test I now set and check the fill_value on the cube rather on the cube.data.fill_value itself (the fill_value on the data is no longer considered by Iris).
• I get/set the field bmdi using the cube.fill_value, regardless of whether or not it is masked data (someone may have real unmasked data but still want a bmdi set)
• When realising a field's data it returns a masked array not an nd array with the nans filled with the bmdi
• I also had to fix the reference text file of the pp file as the repr of a field's data has changed (it used to be this)

[lbdreyer]

Tests are failing because, in those tests, the cube has a fill value of 9999.0 so iris is now setting the bmdi to be that rather than the default bmdi.

The failing tests were

• iris.tests.test_cube_to_pp.TestPPSave.test_pp_append_singles
  ** uses stock.simple_pp() which uses filename = tests.get_data_path(['PP', 'simple_pp', 'global.pp'])

• iris.tests.test_cube_to_pp.TestPPSave.test_pp_save_rules
  ** uses in_filename = tests.get_data_path(('PP', 'simple_pp', 'global.pp'))

• iris.tests.test_cube_to_pp.TestPPSave.test_user_pp_save_rules
  ** uses in_filename = tests.get_data_path(('PP', 'simple_pp', 'global.pp'))

and using iris/master

>>> import iris.fileformats.pp as pp
>>> fields = list(pp.load(<path_to_test_data>/test_data/PP/simple_pp/global.pp'))
>>> print len(fields)
1
>>> print fields[0].bmdi
9999.0

[lbdreyer]

@marqh @corinnebosley Could you take a look at this PR?

I would like this merged in as I have some upcoming PRs based off this work (but makes more sense to be in a separate PR)

[marqh]

@lbdreyer this makes sense to me, based on the updated fill_value handling in the Cube API