fix(pipeline): retry on all LLM errors (not just rate limits)

Free-tier APIs intermittently return invalid JSON or empty responses.
Now any exception in _call_llm retries up to 3 times with a 5s back-off,
rather than failing immediately on non-rate-limit errors.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

markitect/infospace/pipeline.py

@@ -584,9 +584,14 @@ class SourcePipeline:
             except Exception as exc:
                 msg = str(exc)
                 print(f"    LLM error: {msg}")
-                meta["error"] = msg
-                meta["duration_seconds"] = round(time.time() - t0, 1)
-                return None, meta
+                if attempt < max_retries:
+                    wait = 5 * (attempt + 1)
+                    print(f"    Retrying in {wait}s (attempt {attempt + 1}/{max_retries})...")
+                    time.sleep(wait)
+                else:
+                    meta["error"] = msg
+                    meta["duration_seconds"] = round(time.time() - t0, 1)
+                    return None, meta
 
         if response is None:
             meta["error"] = "no response"