Scrub stale NN- prefixes missed by rename sweep

The rename PR's perl sweep keyed on ``examples/NN-name`` paths,
which caught directory references but missed three other forms
of stale numeric prefix:

- Link DISPLAY text in docs/concepts/llms.md (three bullets in
  the "Where to next" section + one inline reference): the link
  targets were updated correctly but the text still read
  ``00 - Hello, world`` / ``07 - Multimodal prompt`` /
  ``09 - Tool use``. Rewriting to drop the prefix and match the
  by-name style used elsewhere.
- Standalone example-name references in source comments (no
  ``examples/`` prefix), e.g. ``00-hello-world`` and
  ``01-routing-and-subgraphs`` in three example main.py files.
  Stripping the leading digit-digit-dash.

Surfaced by CoPilot review on PR #120 (line 338 + 598 + 600
threads), plus a broader git-grep sweep to catch the source-
comment cases the threads didn't flag.

Author: chris-colinsky

docs/concepts/llms.md

@@ -334,8 +334,8 @@ Wiring the loop as a graph cycle: a `call_llm` node, a
 to `call_llm` when `tool_calls` are present and forward to a
 termination node when they aren't. A turn cap on the routing function
 prevents runaway loops on a model that stays in tool-calling forever.
-See [`09 - Tool use`](../examples/tool-use.md) for the runnable
-shape.
+See [the tool-use example](../examples/tool-use.md) for the
+runnable shape.
 
 ### Controlling tool-call behavior with `tool_choice`
 
@@ -591,10 +591,10 @@ classifier won't do this for them.
 - [API reference: `openarmature.llm`](../reference/llm.md) for the
   full surface: message types, `Response`, `RuntimeConfig`, every
   error class, validation helpers.
-- [Examples: 00 - Hello, world](../examples/hello-world.md) for a
+- [Examples: Hello, world](../examples/hello-world.md) for a
   runnable graph exercising both `response_schema` forms in one
   pipeline.
-- [Examples: 09 - Tool use](../examples/tool-use.md) for the
+- [Examples: Tool use](../examples/tool-use.md) for the
   agent-loop pattern with two local tools.
-- [Examples: 07 - Multimodal prompt](../examples/multimodal-prompt.md)
+- [Examples: Multimodal prompt](../examples/multimodal-prompt.md)
   for content blocks alongside versioned prompts.

examples/explicit-subgraph-mapping/main.py

@@ -207,7 +207,7 @@ async def synthesize(s: ComparisonState) -> Mapping[str, Any]:
 #     with no way to express "this site reads A, that site reads B." Both
 #     sites would write `parent.summary` and clobber each other.
 #
-#   - A custom `ProjectionStrategy` (the 01-routing-and-subgraphs approach)
+#   - A custom `ProjectionStrategy` (the routing-and-subgraphs approach)
 #     would have to differ per call site; you'd write two distinct projection
 #     classes that do the same thing in mirror image. That's exactly the
 #     boilerplate `ExplicitMapping` removes.

examples/nested-subgraphs/main.py

@@ -223,7 +223,7 @@ async def pick_doc(s: DocQAState) -> Mapping[str, Any]:
     # return only part of the title. Accept either direction of containment
     # over the lowercased strings; strict equality is too brittle for
     # free-form output. A production app would constrain the model with
-    # response_schema (see 00-hello-world) so the reply is guaranteed to be
+    # response_schema (see hello-world) so the reply is guaranteed to be
     # a valid title.
     match = next(
         (d for d in CORPUS if d["title"].lower() in reply or reply in d["title"].lower()),

examples/routing-and-subgraphs/main.py

@@ -10,7 +10,7 @@
 custom `ProjectionStrategy` for the parent ↔ subgraph boundary, and the
 `merge` reducer for dict accumulation.
 
-Three graph features that `00-hello-world` only touched lightly:
+Three graph features that `hello-world` only touched lightly:
 
   1. **Conditional edges.** The entry node classifies the question and the
      graph routes to one of two branches based on that classification.
@@ -22,7 +22,7 @@
      shape what comes back out), we write a `ProjectionStrategy` by hand.
 
 LLM calls go through ``openarmature.llm.OpenAIProvider`` (same pattern as
-``00-hello-world``) so the example reads as the recommended path rather
+``hello-world``) so the example reads as the recommended path rather
 than as "openai with some openarmature on top."
 
 **Configuration** (env vars; OpenAI defaults shown):