Once attributes and clip paths have IR representations, the complexity of handling different vector asset formats will be deferred and isolated to the reading to and writing from the IR. The format specific conversion implementations should be removed.
This probably shouldn't happen. Different types will still need conversion semantics. There's still a little bit to do in the IR for the conversions to be as simple as possible.
Once attributes and clip paths have IR representations, the complexity of handling different vector asset formats will be deferred and isolated to the reading to and writing from the IR. The format specific conversion implementations should be removed.
In service of #36