ImageBitmap: Change two enum uses.
An earlier CL (https://codereview.chromium.org/1382883002/
) made two changes
which seem to have unintended effects on behaviour.
1. The ImageBitmap constructor asserts that its data is opaque (whereas
previously it took the default NonOpaque opacity mode), but if this is so
it should have a comment justifying it.
2. UnacceleratedImageBufferSurface is clearings the SkSurface if it was
told _not_ to initialize image pixels, which seems like an inversion of the