update the Unfilter API in dsp to process one row independently

This will allow to work in-place on cropped area later.

Also sped up the inverse gradient filtering in SSE2 (~4%)

Change-Id: I463149eee95d36984328f163a1e17f8cabd87441

src/dec/alphai.h

@@ -36,6 +36,7 @@ struct ALPHDecoder {
                        // pixel, sometimes VP8LDecoder may need to allocate
                        // 4 bytes per pixel internally during decode.
   uint8_t* output_;
+  const uint8_t* prev_line_;   // last output row (or NULL)
 };
 
 //------------------------------------------------------------------------------