Wait until A2DP source shutdown completes when active peer is set to the empty address.

Currently, there is a race condition when switching A2DP devices. We
first set the active peer to the empty address and then to the newly
active device. When the active peer is set to the empty address, we
shutdown the A2DP source which takes time. Before that shutdown
completes, we try to set the active peer to the newly active device.
Since the source has not finished shutting down completely, this request
to start a new A2DP source is ignored by the stack.

Fix this by waiting until shutdown is complete and A2DP source state is
set to kStateOff. This way, when setActivePeer with empty address
returns, A2DP source state would already have been set to kStateOff.

Bug: 253974905
Tag: #refactor
Test: manual by using log patterns

Change-Id: I2a51abfd3a68e5efd66dc117fda4b0697816bb04