hifi-decode: Better resampling, Spectral Noise Reduction, Multi processing

[eshaz]

Noise reduction / Re-sampling updates

• Add spectral noise reduction before deemphasis to remove FM noise
• Use sinc converter for resampling in audio chain to reduce resampling artifacts (controlled via options)

Multi-processing updates

• Refactor thread pools into process pools for better multi threading. This helps to offset the extra processing time for the better resampling methods.

Other Updates

• Fix hifi-decode deemphasis and weighting filters to use closer bandwidth to spec, vs. using q factor
• Add ability to specify bandwidth and slope to vhsdecode/addons/FMdeemph.py shelf filter
• Refactor post processing audio chain

New Options

• Add --NR_spectral_amount option to control amount of spectral noise reduction.
  • 0 to disable
  • 1 for full spectral noise reduction
  • default: 0.4
• Add --resampler_quality option to enable better quality resamplers. This reduces noise at the expense of processing time.
  • low, medium, high
  • default: high

[eshaz]

The GUI should probably be checked before this gets merged in.

[eshaz]

Found a small bug, setting back to draft while I fix it.