feat: distributed brain architecture with rqlite and local embeddings (#118)

- Add new brain module with rqlite-based distributed memory and task queue
- Implement BrainClient for memory operations (store, recall, search)
- Implement DistributedWorker for continuous task processing
- Add local embeddings via sentence-transformers (all-MiniLM-L6-v2)
  - No OpenAI dependency, runs 100% local on CPU
  - 384-dim embeddings, 80MB model download
- Deprecate persona system (swarm/personas.py, persona_node.py)
- Deprecate hands system (hands/__init__.py, routes)
- Update marketplace, tools, hands routes for brain integration
- Add sentence-transformers and numpy to dependencies
- All changes backward compatible with deprecation warnings

Co-authored-by: Alexander Payne <apayne@MM.local>

src/brain/__init__.py

@@ -0,0 +1,14 @@
+"""Distributed Brain — Rqlite-based memory and task queue.
+
+A distributed SQLite (rqlite) cluster that runs across all Tailscale devices.
+Provides:
+- Semantic memory with local embeddings
+- Distributed task queue with work stealing
+- Automatic replication and failover
+"""
+
+from brain.client import BrainClient
+from brain.worker import DistributedWorker
+from brain.embeddings import LocalEmbedder
+
+__all__ = ["BrainClient", "DistributedWorker", "LocalEmbedder"]