Two big additions to the voice recognition APIs:

* Allow activities satisfying RecognizerIntent.ACTION_WEB_SEARCH to
  point to the class name of a broadcast receiver which returns details
  about the voice search implementation in an ordered broadcast response.
  Provide a convenience method for getting the intent to fire for this
  info. This can be used to get the current language preference and the
  list of supported languages, and is extensible for future uses.

* When creating a RecognitionManager, allow the caller to optionally specify
  a specific component of a voice recognition service on the device that they
  want to use. This way, an app can still use its own service through
  RecognitionManager, even if it's not the one chosen the user in settings.