feat(recall): wire neural embeddings into the live persona path (#40)

[joelteply]

What

The live persona recall path (build_workspace_cycle → RecallFaculty)
hard-coded the lexical bootstrap embedder. Now it uses neural embeddings (via
unsloth's /v1/embeddings, #1714) when the embed model actually serves, and
falls back to lexical otherwise — personas get semantic recall, not just
word-overlap, the moment an embed model is available. Builds directly on
#1713 (startup auto-fuel) + #1714 (create_embedding).

resolve_recall_embedder(adapter)

• Prefers neural, but probes it once with a one-shot embed. Usable probe
  (non-empty, non-zero, all-finite) → neural; empty/zero/NaN probe (model not
  loaded, endpoint error) → lexical fallback. This is the no-signal guard:
  without it, an adapter that advertises embedding support but has no embed model
  loaded would embed every memory into a zero vector and recall would return
  nothing — strictly worse than lexical.
• Process-stable (decided once at spawn, never per-embed): a query and the
  stored vectors must share one embedding space, so neural and lexical are never
  mixed per call (cosine across spaces is meaningless).
• Wrapped in the content-addressed CachingEmbeddingProvider — each message
  embedded once and shared across every persona (the latency win). Further
  perf is a later pass.
• Always returns a working embedder — never errors/panics. A box with no
  embed model still gets real lexical relevance ("solve for public users").

Wiring

• PersonaBrainConfig gains embedder: Option<Arc<dyn EmbeddingProvider>>;
  build_workspace_cycle uses it, defaulting to lexical+cache when None.
• The live supervisor spawn sets it via resolve_recall_embedder(adapter).
• CANONICAL_EMBED_MODEL = "qwen3-embedding-0.6b" (dim 1024), overridable via
  UNSLOTH_EMBED_MODEL — one embedding space across the grid.

Tests

7 new: the probe gate (usable vs no-signal), and the resolver picking neural when
the model serves / falling back on empty-probe / using lexical when the adapter
lacks embedding support.

Scope

Does not delete fastembed — that path (the separate IPC Hippocampus /
PersonaMemoryManager, real blast radius into ORM vector + search) is the
follow-up trim, deferred per "get it working first, optimize once healthy."

🤖 Generated with Claude Code