Avoid deadlock during R4 call transfer

There was a possibility that during attended call transfer, deadlock
happened in locking bridge_channel and channel.
bridge_channel_queue_deferred_frames() was waiting for brcm_indicate()
to unlock the channel, while brcm_indicate() which called
ast_bridge_channel_queue_frame() was waiting for
bridge_channel_queue_deferred_frames() to unlock the bridge.

Creating new macro to lock both channel and bridge_channel at the same
time and sched_yield() -> retry if failed and use it in
bridge_channel_queue_deferred_frames() avoiding the deadlock.