Implementation of a gaussian downscale using 2 approaches:
1) FastCV
2) Intrinsics

TODO:
1) Develop fastRPC wrappers to allow testing warp on all of native, target and simulator.
2) Add Chroma downscale, not to require deinterleave with 4:2:0 data.
3) Add parameters to make multithreading possible.
4) Perhaps support multiple boundary conditions.