[Accessibility] TTS API interface description + reqs update

[Hynek Hanke]

> > 2.  A say_text() call can speak the text from a specified word or
> > sentence. Would speak from a specified character position also be
> > useful?  Consider an application where the user places a caret in the
> > text and asks to speak from that position onwards.  The application
> > can't ask to speak from a word number, because its counting of words
> > may differ from the synthesizer's.
> The interface is designed with the goal that applications don't have to do 
> their own sentence or word boundary detection.  Given a character position, 
> the application can determine the word boundary by examining the returned 
> word index mark information.

This is not always possible under the current API draft. Consider a
situation when the audio for that part of the text was not generated
yet. Then the application does not have information about the closes
word.

With regards,
Hynek Hanke