Render one assistant message per turn, and make the request timeout a setting

Two setup failures found while calibrating the new benchmark tasks,
both of which would also fail real milestones.

Devstral's chat template (Mistral) rejected any turn in which the model
wrote some text before a tool call: the OpenAI renderer made one message
per event, so the turn became a text message and a tool-call message in
a row, and llama.cpp answered 500. The renderer now folds consecutive
user or assistant messages into one, which is the OpenAI format for a
turn anyway. Checked against the live server: the captured failing
request no longer occurs, over six real runs.

LiteLLMClient gave up on any request after a fixed 120 s. A dense 27B
model writing a whole module in one turn takes longer, so the executor
escalated work that was going well. llm.request_timeout_s now sets it
for both clients, default 600.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

nisse-dev/src/nisse_dev/agent.py

@@ -144,7 +144,10 @@ def make_llm_clients(
 
     normalized = budgeted(
         LiteLLMClient(
-            config.llm_base_url, config.llm_token, max_tokens=config.max_output_tokens
+            config.llm_base_url,
+            config.llm_token,
+            max_tokens=config.max_output_tokens,
+            timeout_s=config.request_timeout_s,
         )
     )
     # Delegations prefer the vendor-native route; fidelity is cheap
@@ -152,7 +155,10 @@ def make_llm_clients(
     # select_client, not by where the caller happens to sit.
     passthrough = budgeted(
         AnthropicPassthroughClient(
-            config.llm_base_url, config.llm_token, max_tokens=config.max_output_tokens
+            config.llm_base_url,
+            config.llm_token,
+            max_tokens=config.max_output_tokens,
+            timeout_s=config.request_timeout_s,
         )
     )
     return normalized, passthrough