NiNode flags now default to 8 for "Not a skin influence."

The code I added to make NiTriBasedGeom select the correct skeleton root was wrong and has been re-written and tested.  It now seems to select the same skeleton root as real NIF files.
The NiTriBasedGeomData::SetUVSetCount function now sets the hasUVs bool value correctly.
Fixed some skin related bugs in ComplexSh::Split.