Kill Topic Classifier & Refactor Digest Worker

The TopicClassifierService, TwoSignalBoundaryService, and RecentTopicService were completely removed, along with all associated tests and benchmarks.

In backend/workers/digest_worker.py, several imports were removed, and the main path logic was drastically simplified to use thread_id as the context topic.

Classification logic in digest_worker.py was replaced with a static synthetic result and a direct call to EmbeddingService().generate_embedding(text).

Several counters and timing metrics in digest_worker.py, thread_conversation_service.py, and metrics_service.py were renamed from ‘classification’ to ‘embedding’ for consistency.

The TopicContext dataclass in backend/services/topic_context.py was cleaned up by removing classifier-specific fields like is_boundary and just_reset_from_silence.

Dead code cleanup included removing stale Redis key patterns referencing the deleted services from backend/api/privacy.py.

The schema service was updated in backend/services/schema_service.py by removing the topics_vec virtual table reference.

Stale benchmark and results artifacts related to the boundary service were deleted from backend/tests/benchmarks/results/.

• Deleted TopicClassifierService, TwoSignalBoundaryService, RecentTopicService and related code.

• Digest worker now uses direct EmbeddingService + thread_id as topic key.

• Metrics counters and operations were renamed from ‘classification’ to ‘embedding’.

• Removed classifier-specific fields from TopicContext.

• Stale Redis key patterns referencing the removed services were cleaned up.

• The topics_vec virtual table was removed from schema definitions.