WeSinger: Data-augmented Singing Voice Synthesis with Auxiliary Losses