[voiceinteraction] Personal Assistant Architecture draft with use case

I started a draft of an architecture document at
https://github.com/w3c/voiceinteraction/tree/master/voice%20interaction%20dr
afts and I put in a simplified version of the use case I had sent earlier.
The old use case made some assumptions about the architecture. I removed
those because a use case should really be free from any assumptions about
how it should be addressed. The next step is for Dirk to add the
architectural information from his pdf.