Implement music, sound, and positional-audio presentation

[zoeyrose]

Outcome

Deliver robust SDL3 audio presentation from separately licensed sound resources without coupling audio to GPU rendering.

Scope and invariants

• Own audio device open/change/loss, decode/mix/playback queues, category volume/mute, music transitions, positional attenuation/pan where approved, and application focus behavior.
• Consume verified resource-provider entries and semantic server/session sound events; never accept executable codecs/plugins from a server.
• Define overlapping-instance, priority/voice limits, looping, cancellation, map/session transition, reconnect, and shutdown behavior.
• Preserve asset-specific licenses/attribution in packaging and diagnostics.
• Provide a fake audio backend/event recorder for headless tests.

Acceptance criteria

• Tests cover missing/corrupt/oversized files, device hotplug/loss, rapid transitions, duplicate/stale events, voice limit, mute/volume migration, reconnect, and teardown.
• No sound leaks indefinitely after entity/map/session destruction.
• Audio queues/memory/voices are bounded and observable.
• M3 plays representative UI/world/music events through approved assets and still functions with audio disabled/unavailable.
• Renderer has no audio dependency.

Dependencies and parallelization

Depends on client platform #4, resource #32, and sound provenance/release contracts. Can develop independently of world rendering.

Licensing

Client code/tests are MIT. Legacy reuse requires an audited approved-grantor grant; all other implementation is independent. Content, sound, and graphics retain their exact asset licenses and notices.