fix(notebooks): use full UUID for run tags; drop stub schema methods from embedded adapter

- 01_llama, 02_langchain: use full uuid4().hex for `tag` variable
  (was [:6] = 24 bits entropy, risked cross-run collisions on shared envs)
- 01_llama, 02_langchain: remove get_labels()/get_edge_types() from
  _EmbeddedAdapter — returning [] made adapter look schema-RPC-capable,
  silently short-circuiting the get_schema_text() path in CoordinodeGraph
- 03_langgraph: replace explicit LIMIT N > 20 in query_facts via re.sub
  instead of only appending when absent; server no longer materializes
  oversized result sets from LLM-generated queries

fix(client): remove contradictory "indexed automatically" claim from
text_search() Args.label docstring; Note already says indexing requires
explicit CREATE TEXT INDEX DDL

Author: polaz

coordinode/coordinode/client.py

@@ -686,10 +686,7 @@ async def text_search(
         """Run a full-text BM25 search over all indexed text properties for *label*.
 
         Args:
-            label: Node label to search (e.g. ``"Article"``). On current
-                CoordiNode servers, text properties in schema-free graphs are
-                indexed automatically; labels with no searchable text content
-                simply return ``[]``.
+            label: Node label to search (e.g. ``"Article"``).
             query: Full-text query string. Supports boolean operators (``AND``,
                 ``OR``, ``NOT``), phrase search (``"exact phrase"``), prefix
                 wildcards (``term*``), and per-term boosting (``term^N``).