fix(llm): honor LLM_TIMEOUT during scans, not just warm-up

LLM_TIMEOUT is documented (docs/advanced/configuration.mdx) as the
request timeout for LLM calls and is offered as the workaround for slow
local models, but it was only applied to the warm-up call in
strix/interface/main.py. Scan LLM calls go through the SDK's LiteLLM
model, which invokes litellm.acompletion without a timeout and falls
back to LiteLLM's module default (6000s) -- so `export LLM_TIMEOUT=600`
had no effect on the actual run.

Set litellm.request_timeout from settings in configure_sdk_model_defaults
so the documented setting takes effect for LiteLLM-routed scans.

Fixes #426

strix/config/models.py

@@ -63,6 +63,7 @@ def configure_sdk_model_defaults(settings: Settings) -> None:
     llm = settings.llm
     set_tracing_disabled(True)
     _configure_litellm_compatibility()
+    _configure_litellm_request_timeout(llm.timeout)
     if llm.api_key:
         set_default_openai_key(llm.api_key, use_for_tracing=False)
         _configure_litellm_default("api_key", llm.api_key)
@@ -128,6 +129,23 @@ def _configure_litellm_default(name: str, value: str) -> None:
     setattr(litellm, name, value)
 
 
+def _configure_litellm_request_timeout(timeout: int) -> None:
+    """Apply the configured ``LLM_TIMEOUT`` to LiteLLM-routed scan calls.
+
+    The SDK's LiteLLM model invokes ``litellm.acompletion`` without an explicit
+    per-request timeout, so without this it falls back to LiteLLM's module
+    default and the documented ``LLM_TIMEOUT`` only affected the warm-up call
+    (``strix/interface/main.py``). Setting the module-level default makes
+    ``export LLM_TIMEOUT=600`` take effect for the actual scan, restoring the
+    documented behavior for slow local / self-hosted models.
+    """
+    import litellm
+
+    # litellm doesn't re-export request_timeout in its public surface, but it is
+    # the module-level default it reads for each acompletion call.
+    litellm.request_timeout = timeout  # type: ignore[attr-defined]
+
+
 def uses_chat_completions_tool_schema(model_name: str, settings: Settings) -> bool:
     """Return whether the resolved SDK route can only receive JSON function tools."""
     model = model_name.strip().lower()