Fix GoldParse init

[svlandeg]

The GoldParse init constructor has a catch-all **_ which seems to have led to a subtle bug in nn_parser.pyx.

Personally I would prefer removing **_ from the constructor, but this may break people's scripts (though the breaking might be informational). We can also remove it as part of the v.3.0 release. Unless there is a reason to keep this?

Types of change

bug fix

Checklist

• I have submitted the spaCy Contributor Agreement.
• I ran the tests, and all new and existing tests passed.
• My changes don't require a change to the documentation, or if they do, I've added all required information.

[honnibal]

I think we had the catch-all there for some backwards compatibility at some point. Removing it seems fine.