Add denoiser to preprocessing step

Marc and Tim both report large gains after denoising. We should incorporate https://github.com/roedoejet/denoiser into EveryVoice and allow it to be configured to run during the preprocessing step.

Start by locating EveryVoice's preprocessing entry point and configuration mechanism, then read the linked denoiser project's integration guidance. Done means denoising can be enabled and configured during preprocessing without changing the existing default behavior; no file or test is named in the issue.

Written by the indexing model from the issue text.