H264: Only wait before triggering ff_thread_setup_complete() until the next slice that contains a start-of-field/frame macroblock

This allows concurrent decoding of the last field/frame, rather than
only the last slice, of data packets with multiple NAL units packed
together.

This will fix the slowdown reported in e.g. bug 52.
Signed-off-by: Anton Khirnov <anton@khirnov.net>