rv40: make rv40_adaptive_loop_filter() always_inline

This function is called from the rv40_[hv]_loop_filter() wrappers
with some arguments constant, so it really needs to be inlined.

3% faster overall on Cortex-A8.

Originally committed as revision 23893 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/rv40.c

@@ -307,7 +307,7 @@ static inline void rv40_weak_loop_filter(uint8_t *src, const int step,
     }
 }
 
-static inline void rv40_adaptive_loop_filter(uint8_t *src, const int step,
+static av_always_inline void rv40_adaptive_loop_filter(uint8_t *src, const int step,
                                              const int stride, const int dmode,
                                              const int lim_q1, const int lim_p1,
                                              const int alpha,