Add audio steganography with LSB and spread spectrum modes

Implement two audio embedding modes following the same multi-factor
authentication pipeline as image steganography (passphrase + PIN +
optional RSA key + optional channel key):

- audio_lsb: High-capacity LSB embedding in PCM samples for lossless
  formats (WAV/FLAC). Uses ChaCha20-keyed sample index selection.
- audio_spread: Direct-sequence spread spectrum (DSSS) with ChaCha20-
  keyed bipolar chip codes, Reed-Solomon error correction, and 3-copy
  majority-voted length headers. Designed to survive lossy compression.

New files:
- audio_steganography.py: LSB embed/extract on PCM samples
- spread_steganography.py: Spread spectrum embed/extract
- audio_utils.py: Format detection, transcoding, validation helpers
- tests/test_audio.py: 22 tests covering both modes end-to-end

Updated encode.py, decode.py, cli.py (audio-encode/audio-decode
commands), constants.py, models.py, exceptions.py, validation.py,
__init__.py, and pyproject.toml ([audio] extra).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

pyproject.toml

@@ -52,6 +52,13 @@ dct = [
     "jpeglib>=1.0.0",
     "reedsolo>=1.7.0",
 ]
+audio = [
+    "pydub>=0.25.0",
+    "numpy>=2.0.0",
+    "scipy>=1.10.0",
+    "soundfile>=0.12.0",
+    "reedsolo>=1.7.0",
+]
 cli = [
     "click>=8.0.0",
     "qrcode>=7.30",
@@ -86,7 +93,7 @@ api = [
     "reedsolo>=1.7.0",
 ]
 all = [
-    "stegasoo[cli,web,api,dct,compression]",
+    "stegasoo[cli,web,api,dct,audio,compression]",
 ]
 dev = [
     "stegasoo[all]",
@@ -141,6 +148,8 @@ ignore = ["E501"]
 [tool.ruff.lint.per-file-ignores]
 # YCbCr colorspace variables (R, G, B, Y, Cb, Cr) are standard names
 "src/stegasoo/dct_steganography.py" = ["N803", "N806"]
+# MDCT transform variables (N, X) are standard mathematical names
+"src/stegasoo/spread_steganography.py" = ["N803", "N806"]
 # Package __init__.py has imports after try/except and aliases - intentional structure
 "src/stegasoo/__init__.py" = ["E402"]