SSML Prosody: Support duration attribute

[earboxer]

https://www.w3.org/TR/speech-synthesis11/#edef_prosody

duration: a value in seconds or milliseconds for the desired time to take to read the contained text. Follows the time value format from the Cascading Style Sheet Level 2 Recommendation [CSS2], e.g. "250ms", "3s".

I believe this would take considerable work, as the SSML parser may not handle encapsulation properly.

I don't think our SSML parser supports unit-detection, so I would suggest using milliseconds as the default (presumed) unit.