audio: enable concurrent capture for voice recognition

Add AUDIO_SOURCE_VOICE_RECOGNITION to the list of use
cases allowed for concurrent capture. As AUDIO_SOURCE_HOTWORD is
enabled for concurrent capture it makes sense to enable a client
to transition to voice recognition after the hotword has been detected
in a concurrent capture situation.

Also fixed some concurrent capture scenarii

Bug: 22702906
Test: build and check "OK Google" detection
Change-Id: Ic6dd41372fe74122395764beca7ca9a65197464b