fixed bugs in fastpitch tts synthesis

…2587)

* It looks like the Neon model is special in that t does not have a speaker_name and it wants to get the only item available. This was blocking a valid model with one speaker and a d_vector_file from being executed to get the embedding.

* Update synthesizer.py

oh my how embarrassing

* add add cli options for play and speed
--play argument uses simpleaudio to play the tts wav
--speed <float 0.0-2.0> passes speed argument to Coqui Studio models

* remove simpleaudio not referenced in file

* fix simpleaudio dependency version

* add ALSA headers for simpleaudio compilation

* Dockerfile ALSA headers for simpleaudio

* base changes to use stdout instead of play audio
Considering conversion to pipe wav data for audio playback with ohter program
like aplay.

This is incomplete code. Using to get feedback before proceeding with
implementation.

* remove play for pipe_out arg that suppresses stdout
removed play and simpleaudio dependency in place of pipe
fuctionality to allow passing wav file data to a program
dedicated to playing audio.

* scipy.io.wavfile.write fails with /dev/null target

* Streaming inference for XTTS 🚀 (#3035)

* v0.17.7

* Redownload XTTS with the local and remote config do not match

* Remove unused method

* Print a message when it is already donwloaded

* Try-except to present error when the user dont have connection

* Fix style

* 0.17.8

* v0.17.8

---------

Co-authored-by: Julian Weber <[email protected]>
Co-authored-by: Eren Gölge <[email protected]>
Co-authored-by: Edresson Casanova <[email protected]>
Co-authored-by: ggoknar <[email protected]>

* fix formatting dataset doc

* fix autocomplete

* fix readme

* fix inference.md