Add new TTS API.

Much of existing API is reused (mostly on the service side). The new API
provides better support for network-based synthesis, defines explicit
synthesis fallback support, has better error handling, allows for multiple
different voices per locale, features enumeration and custom extensions
on a voice level and many more.

If a service does not provide a V2 implementation, a basic adapter
from V2 to existing V1 implementation is provided.

Bug: 8259486
Change-Id: I797b7fc054f11c9b0d930f4dea582e57596ea127