feat(config): expose router generation + narration knobs in config file

The router/narration model behaviour was hardcoded (Main built RouterConfig()
with defaults). Surface it under the [router] section so it's tunable without a
rebuild:

  [router]            conversation_keep_last, retrieval_k, token_budget
  [router.generation] temperature, top_p, max_tokens          (chat/steering)
  [router.narration]  temperature, top_p, max_tokens, max_per_run

ConfigLoader parses the new sections (adds asDouble); Main maps the config-layer
RouterConfig onto the domain RouterConfig and threads narration.max_per_run into
ServerModule. All values default to the previous constants, so behaviour is
unchanged when the sections are absent. Documents the block in sample-config.toml
and adds parser tests for present/absent cases.

docs/sample-config.toml

@@ -74,6 +74,26 @@ capabilities = { General = 1.0, Coding = 0.7, Reasoning = 0.6, Summarization = 0
 # capabilities = { General = 1.0, Coding = 0.7, Reasoning = 0.6, Summarization = 0.8, ToolCalling = 0.5 }
 
 # Router configuration (optional, defaults shown below)
+[router]
+conversation_keep_last = 6     # how many recent chat turns to keep in context
+retrieval_k = 5                # L3 memory hits to retrieve per query
+token_budget = 4096            # context budget for router prompts
+
+# Sampling/length for router chat + steering replies.
+[router.generation]
+temperature = 0.7
+top_p = 0.9
+max_tokens = 512
+
+# Router narration (the live workflow commentary). max_tokens is higher than chat so
+# reasoning models can finish "thinking" and still emit the line; max_per_run caps how
+# many narrations fire per workflow run.
+[router.narration]
+temperature = 0.7
+top_p = 0.9
+max_tokens = 1024
+max_per_run = 100
+
 [router.embedder]
 backend = "noop"               # or "llamacpp"
 dimension = 1536