Also by wrapping the SHM buffer in an AVBufferRef we eliminate yet another
possible memcpy improving performance.
Signed-off-by: Marton Balint <cus@passwd.hu>