required __init__ args currently fail on wrapping

[dalejung]

Support init params for things like series + 1. While metadata propogates, currently (2013/07/01) wrapping fails because it calls the constructor instead of calling .view

        class SubSeries(UserSeries):
            def __init__(self, *args, **kwargs):
                bob = kwargs.pop('bob')
                self.bob = bob
                super(SubSeries, self).__init__(*args, **kwargs)

        ss = SubSeries(range(10), bob=123)
        assert ss.bob == 123
        test = ss + 1 # currently errors
        assert test.bob == 123

[dalejung]

Fixed by passing in the meta dictionary as keyword args to the subclass constructor. It's possible that most of these args aren't used by the constructor, which shouldn't be a problem as long as they don't clash with any DataFrame/Series constructor args.

There is an _init_args check that makes sure to only grab the appropriate arguments when calling the pandas constructors.