speech demos issues

[pkra]

As discussed with @zorkow.

Besides #41 there are other issues with the demos

• container is stripped by default - but the container has the aria-label
• the container lacks a suitable role - custom elements have role=generic which does not allow for aria-label. With a flat label approach of these demos, I tend to go with img for simplicity's sake.

I also had issues with xml validation but I don't recall if that was a problem with the demo or with the jsdom/litedom adaptors.

[zorkow]

#49 should fix these things. Even for deep speech you get aria-label and role attributes. If one were to add a tabindex=0 one could even do some rudimentary exploration, similar to hichart diagrams.

But just as a warning: These demos should show how to play with speech, not taken as fully developed a11y solutions.