audio VTS: Refactor support for non-deterministic SM behavior

Use a DAG instead of a list for representing stream state machine
transitions. This allows accomodating multiple possible replies
from the stream state machine, and enables following different,
possibly converging test sequences depending on actual replies.

Update the async output state machine graph including possible
state transitions that are implemented but were missing from the
graph.

Bug: 262402957
Test: atest VtsHalAudioCoreTargetTest
Change-Id: Ie97f3ce9222eec812d4eb4dfbce1f678370bef26