Allow for synthesize to be called again for the same wav_file

[ygerlach]

Currently you can not call synthesize multiple times for the same wav file. So doing stuff like creating some tts with model a and then appending some tts with model b is not possible, because setting wav parameter, after some frames where written is not allowed (throws a wave.Error)

This change makes sure to catch the possible error, but still ensures the parameter in wav_file are still correctly set for the used model.

[ygerlach]

@synesthesiam can you take a look at this PR, please?

[synesthesiam]

A set_wav_format boolean has been added to synthesize_wav here: https://github.com/OHF-Voice/piper1-gpl