Auphonic batch concatenation: avoid 3-minute minimum billing

gistfile0.txt

# Auphonic Batch Concatenation

## Problem
91% of Auphonic productions were hitting the 3-minute minimum billing floor,
wasting ~80% of credits on short audio clips (typically 3-60 seconds each).

## Solution
Added batch concatenation to `auphonic_automation.py`:
- Accumulates short WAV files arriving in `~/Dropbox/Auphonic/IN/`
- Interactive macOS dialogs show batch status and offer Wait/Send/Cancel
- On Send, prompts for processing settings (default: speech_isolation N9 V9 B9)
- Concatenates files with 1s silence gaps via ffmpeg (normalizes to 24-bit PCM)
- Uploads single concatenated file to Auphonic API (1 production instead of N)
- Splits processed result back into individual files and delivers to OUT/

## Key details
- Threshold: 170s (under 3 min with margin)
- Handles both 24-bit PCM and 32-bit IEEE float WAVs
- Manual WAV header parser as fallback for IEEE float (Python `wave` module can't read them)
- Startup cleanup of stale batch state and processing markers
- REAPER Lua scripts updated: removed settings menu (now handled by daemon), added `set_action_options(1)` to prevent "terminate instances" dialog

## Billing investigation
- Auphonic refunded 22h recurring + 10h onetime credits after a looping bug burned credits
- The 22h recurring credits expired at monthly billing cycle (~March 11)
- User also bought 2 onetime packs (10h + 5h = $34), but only 10h was refunded
- With batching, effective credit usage drops ~5x for short clip workflows

## Commit
`f790d36` on mumblequatch/scripts main branch