diffsinger renderer: stricter vocoder-acoustic compatibility check

[oxygen-dioxide]

• Vocoder's hop size and sample rate should be equal to that of the acoustic model. If they aren't the same, openutau will show an error dialog.
  • Hop size is 512 by default. Sample rate is 44100 by default. If you trained a model with a different hop size or sample rate, you should specify in dsconfig.yaml
• Variance model can have a different hop size and sample rate than acoustic model. If they're different, openutau will resample the variance curves.
• The shape of vocoder output should be (1, length). If it isn't, openutau will show an error dialog.