feat(room_io): mix audio from every participant into one AgentSession

[Darshak03]

Closes #6795

Problem

AgentSession can only listen to one participant at a time: _ParticipantInputStream keeps a single _stream/_publication, so a second participant's track replaces the first instead of joining it. Use cases like an AI interview with human takeover need N participants → 1 session → 1 shared chat context.

Approach

Opt-in mixing. With AudioInputOptions.mix_participants=True, RoomIO subscribes to the microphone of every accepted participant and mixes them with rtc.AudioMixer (the same primitive BackgroundAudio uses) into the single audio input the session already consumes. One STT/LLM/TTS pipeline, one chat context, no API surface beyond the flag:

session.start(
    room=ctx.room,
    room_options=RoomOptions(audio_input=AudioInputOptions(mix_participants=True)),
)

Changes

• voice/room_io/_input.py — _ParticipantAudioInputStream keeps a _MixedSource per participant (channel + track stream + forward task) and feeds each channel into an rtc.AudioMixer; the mixed output becomes the session's input. The only change to the shared single-participant path is a _sink(participant) hook in the base class, which still returns _data_ch — video and non-mixed audio behave exactly as before.
• voice/room_io/room_io.py — every accepted participant is added to the mix on connect and removed on disconnect. Kind filtering is unchanged, and the agent's own avatar worker (ATTRIBUTE_PUBLISH_ON_BEHALF) is excluded so the agent can't hear itself. The linked participant is still the first one and only drives the outputs (audio, transcription, chat text).
• voice/room_io/types.py — the mix_participants option.

Notes:

• AGC runs once on the mixed output rather than per stream (one shared AudioProcessingModule interleaved across speakers would be wrong).
• A noise-cancellation selector now builds a processor per participant, owned by that participant's stream; a directly-passed FrameProcessor instance keeps today's shared lifetime.
• set_participant() is a no-op for the input in this mode — listening covers everyone; toggle it with the existing audio-input enable/disable.

Not included

Per-turn speaker attribution, also mentioned in the issue. Mixing collapses everyone into one stream, so a single STT cannot label who spoke — that needs per-participant recognition, which is a separate change.

Tests

tests/test_room_io.py:

• test_mix_participants_sums_every_participant_audio — two participants' frames come out of the input summed into one frame.
• test_mix_participants_tracks_room_membership — RoomIO mixes both humans, skips its own avatar worker, links only the first, and drops a participant on disconnect.

uv run pytest --unit passes (the pre-existing test_sent_tokenizer failure and test_room.py errors are environment-related and reproduce on main).