fix(matrix): accept is_voice kwarg in send_voice to match base class contract

[hermes-gadget]

Summary

The base class delivery loop passes is_voice= to adapter.send_voice(), but the Matrix adapter's send_voice() did not accept that keyword argument. This caused a TypeError that silently dropped all audio/TTS attachments on Matrix.

Root Cause

Gateway logs showed:

[Matrix] Delivering 1 non-image MEDIA attachment(s)
[Matrix] Error sending media: MatrixAdapter.send_voice() got an unexpected keyword argument 'is_voice'

The base class at gateway/run.py:24328 calls:

await adapter.send_voice(
    chat_id=event.source.chat_id,
    audio_path=media_path,
    metadata=_thread_meta,
    is_voice=is_voice,
)

But the Matrix adapter's send_voice() signature lacked the is_voice parameter.

Fix

Added is_voice: bool = False to the Matrix adapter's send_voice() signature. The parameter is unused internally since the Matrix adapter handles voice transcoding (to Ogg/Opus for MSC3245) independently — it just needs to accept the kwarg without crashing.

Test Plan

• Verified TTS voice notes now deliver correctly on Matrix after the fix
• Voice notes render as voice bubbles (MSC3245) in Element
• Gateway restart required to pick up the plugin change

[n30nex]

🦀 Krab review: Nice root-cause fix—small, contract-correct, and directly tied to the observed TypeError. One worthwhile follow-up would be a focused regression test proving the delivery loop can call the Matrix adapter with is_voice= without raising, so a future signature drift cannot silently drop voice attachments again. Otherwise: lean fix, ship it.

[hermes-gadget]

Rebased onto current main after #102117

This PR was originally opened against the pre-refactor codebase. After the massive codebase simplification in #102117 (436K additions / 784K deletions across 2,655 files), the old branch had thousands of conflicts and was stale.

What changed:
Recreated the fix on a fresh branch off current upstream/main. The diff is identical in intent — same 1-line addition to plugins/platforms/matrix/adapter.py:

async def send_voice(
    self, chat_id: str, audio_path: str, caption: Optional[str] = None, reply_to: Optional[str] = None,
    metadata: Optional[Dict[str, Any]] = None,
    is_voice: bool = False) -> SendResult:

Why the fix is still needed:
The refactor decomposed gateway/run.py into many smaller modules, but the base class _deliver_media method (gateway/platforms/base.py:3690) still calls:

result = await self.send_voice(chat_id=chat_id, audio_path=path, metadata=metadata, is_voice=is_voice)

The Matrix adapter inherits _deliver_media (doesn't override it) but its send_voice() override lacks the is_voice parameter, so the TypeError and silent audio drop still occur without this fix.