Only run subset of tests from setup.py test

The setuptools invocation of the tests does not put
the test metadata in place, so almost all of the tests
fail.  Only run a few, real metadata-based, tests 
instead.

python/setup.py

@@ -64,7 +64,7 @@
                      url='https://github.com/daviddrysdale/python-phonenumbers',
                      license='Apache License 2.0',
                      packages=pkgs,
-                     test_suite="tests",
+                     test_suite="tests.examplenumberstest",
                      platforms='Posix; MacOS X; Windows',
                      classifiers=['Development Status :: 5 - Production/Stable',
                                   'Intended Audience :: Developers',

setup.py

@@ -45,7 +45,7 @@
                      packages=['phonenumbers', 'phonenumbers.data', 'phonenumbers.geodata', 'phonenumbers.shortdata',
                                'phonenumbers.carrierdata', 'phonenumbers.tzdata'],
                      package_dir={'': 'python'},
-                     test_suite="tests",
+                     test_suite="tests.examplenumberstest",
                      platforms='Posix; MacOS X; Windows',
                      classifiers=['Development Status :: 5 - Production/Stable',
                                   'Intended Audience :: Developers',