Improve audio normalization for voice clarity

[phoenixray2000]

Summary

• replace peak normalization with voice-oriented loudness normalization and limiting
• normalize mic and loopback sources before mixing in both mode
• add regression coverage for spikes, stereo balance, and existing output formats

Stack

• Depends on upstream PR Add output formats and quality controls #4: Add output formats and quality controls
• This PR uses fork branch phoenixray2000:codex-normalize-audio-upstream because this account cannot push stack base branches to lukmay/QuickAudioRecorder.

Validation

• python -m unittest discover -s tests
• git diff --check origin/codex-output-format-quality-stereo-upstream..codex-normalize-audio-upstream