Improve text to speech

Add a pull down menu in General to let the user select which voice to use.

In a separate commit, add a dependency to rubband, a time stretcher. It uses the same API as the C++ Rubber Band library.

In another separate commit, we are going to render by phrases and then stretch time by words.

• Split the text into phrases, by this punctuation: .:;!?.
• For each phrase:
  • Go through the first rendering pass as before
  • In the second rendering pass, use the end of word callback from the text to speech library to establish the beginning and end of each word
  • Use rubband to timestretch each word individually so it fits into the length of the word in the note synthesizer
  • Align the start of each phrase with the time of start that phrase in the note synthesizer, but do not stretch.

Start by locating the General settings UI and the existing text-to-speech rendering path; the issue does not name files or tests. Review how the current two rendering passes and timing callbacks work, then check the rubband dependency API. Done means voice selection is available and phrase/word rendering follows the specified timing behavior.