feat: add RAG plugin, chat completion provider, and resolve-time embedding

- Add chat.ts: chat completion provider abstraction (Ollama via @agentic-kit/ollama)
- Add rag-plugin.ts: LlmRagPlugin with ragQuery and embedText root query fields
  - Detects @hasChunks smart tags to discover chunk tables
  - RAG flow: embed prompt → search chunks → assemble context → chat LLM → return answer
  - Adds RagResponse, RagSource, EmbedTextResponse types to schema
- Wire resolve-time embedding in text-search-plugin (resolver wrapper on queries)
- Wire resolve-time embedding in text-mutation-plugin (resolver wrapper on mutations)
- Update llm-module-plugin to resolve chat completer on build
- Add enableRag toggle to GraphileLlmPreset (default false)
- Add chat completion types (ChatFunction, ChatConfig, ChatMessage, etc.)
- Add RagDefaults, ChunkTableInfo types
- Add integration tests: RAG schema enrichment, ragQuery execution, embedText,
  chat completion abstraction unit tests, preset toggle tests
- Add articles_chunks seed table for RAG testing

Author: pyramation

graphile/graphile-llm/package.json

@@ -38,13 +38,17 @@
     "graphile-build-pg": "5.0.0",
     "graphile-config": "1.0.0",
     "graphile-search": "workspace:^",
+    "graphile-utils": "5.0.0",
     "graphql": "16.13.0",
     "pg-sql2": "5.0.0",
     "postgraphile": "5.0.0"
   },
   "peerDependenciesMeta": {
     "graphile-search": {
       "optional": true
+    },
+    "graphile-utils": {
+      "optional": true
     }
   },
   "devDependencies": {