dixie: add normal loopfilter

This is the more naive implementation as described in the bitstream guide,
rather than the masking version implemented in the reference code. However,
the core function prototypes were left as-is to make it easy to plug in the
reference assembly code.

Verified loopfiltered output matches reference decoder for 500 frames.

Change-Id: Ib4f197e864f07dbb918b6d5e742c6110d57c1f40