Behaviour definition for neuroevolution lineage event persistence.

This behaviour defines the minimal API for lineage tracking. Implementations MUST be non-blocking to avoid impacting evolution performance.

Callbacks

init/1 - Initialize backend persist_event/2 - Fire-and-forget single event persist_batch/2 - Fire-and-forget batch of events read_stream/3 - Read events (for recovery/replay) subscribe/3 - Subscribe to stream (for projections) unsubscribe/3 - Unsubscribe from stream

Performance Requirements

Lineage tracking must NEVER block the evolution loop:

- persist_event/persist_batch should return immediately - Use async I/O internally (spawn, cast, buffering) - Acceptable to lose events under extreme load - read_stream may block (only used for recovery)

Stream Design

Events are organized into streams based on entity type:

- individual-{id} : Birth, death, fitness, mutations - species-{id} : Speciation, lineage events - population-{id} : Generation, capacity events - coalition-{id} : Coalition lifecycle