feat(brain): add brain_answer and brain_classify MCP tools

Adds two new LLM-backed MCP tools to the ingestion service:

- brain_answer(query): BM25 retrieval + LLM synthesis → answer + sources
- brain_classify(text): classifies doc into type/title/tags via LLM

Adds llm.Router for primary→fallback routing (berget.ai → iguana).
Wired via BRAIN_LLM_PRIMARY_URL/BRAIN_LLM_FALLBACK_URL env vars;
no-op when unset so existing deployments are unaffected.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

ingestion/internal/mcp/handlers.go

@@ -69,6 +69,20 @@ func (s *Server) tools() []map[string]any {
 				"dry_run": map[string]any{"type": "boolean"},
 			}),
 		},
+		{
+			"name":        "brain_answer",
+			"description": "Retrieve relevant brain content via BM25 and synthesize a coherent answer using an LLM.",
+			"inputSchema": schema([]string{"query"}, map[string]any{
+				"query": str("question to answer"),
+			}),
+		},
+		{
+			"name":        "brain_classify",
+			"description": "Classify raw text into doc type, title, and tags using an LLM.",
+			"inputSchema": schema([]string{"text"}, map[string]any{
+				"text": str("raw document text to classify (first 3000 chars used)"),
+			}),
+		},
 		{
 			"name":        "session_log",
 			"description": "Append a structured entry to brain/sessions/<session_id>.jsonl.",