Some answers:
1) You can have multiple bound boxes in the Spine project and C3 can differentiate between multiple bounding boxes at runtime. (The expression for the bounding box center and for the poly points use slot and bounding box name.)
2) The bounding boxes in Spine just provide position and poly data in C3, for hit box and collision, you would still need to align some other sprites with either the center position or use a mesh to have it follow the poly points. I don't think you can get away without having _some_ sprite along with the Spine object. It has to do how the Spine object is rendered, it's basically rendered into a texture which is then used as texture for a rectangular C3 drawing object, so C3 is not aware of all the pieces being rendered. Check out the example project on github for the image you showed below, it shows how to use the poly point of a bounding box to deform a C3 sprite.
Performance wise, using just the bbox center and a C3 sprite will be faster, because you can just set the position of the sprite in one action. To match a changing bbox (e.g. deform) with a sprite w/ mesh points will require more perf due to C3 calculations of bbox, etc. when mesh points change.