Optimize theora decoding with msa in functions: 1. ff_vp3_idct_add_msa 2. ff_vp3_idct_put_msa 3. ff_vp3_idct_dc_add_msa 4. ff_vp3_v_loop_filter_msa 5. ff_vp3_h_loop_filter_msa 6. ff_put_no_rnd_pixels_l2_msa Theora decoding speed improved about 36%(from 22fps to 30fps, Tested on loongson 2K1000). Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>