feat: custom judge criteria and multi-dimensional scoring

Add configurable judge rubrics and multi-dimensional scoring with
weighted aggregation. New config fields: judge_criteria (free text)
and judge_dimensions (list of {name, weight, description}). CLI
--judge-criteria flag provides quick overrides. The judge adapter
computes weighted aggregate scores and enriches feedback with
per-dimension breakdowns.

Co-Authored-By: Paperclip <noreply@paperclip.ing>

tests/unit/test_error_handling.py

@@ -124,12 +124,11 @@ class TestCircuitBreaker:
             circuit_breaker_threshold=3,
             error_strategy="skip",
         )
-        with patch.object(loop, "_log"):
-            state = await loop.run(
-                Prompt("test"),
-                [SyntheticExample("in", id=0), SyntheticExample("in2", id=1)],
-                "task",
-            )
+        state = await loop.run(
+            Prompt("test"),
+            [SyntheticExample("in", id=0), SyntheticExample("in2", id=1)],
+            "task",
+        )
 
         error_events = [h for h in state.history if h.get("event") == "error"]
         cb_events = [h for h in state.history if h.get("event") == "circuit_breaker"]
@@ -165,13 +164,12 @@ class TestCircuitBreaker:
             circuit_breaker_threshold=3,
             error_strategy="abort",
         )
-        with patch.object(loop, "_log"):
-            with pytest.raises(RuntimeError, match="LLM down"):
-                await loop.run(
-                    Prompt("test"),
-                    [SyntheticExample("in", id=0), SyntheticExample("in2", id=1)],
-                    "task",
-                )
+        with pytest.raises(RuntimeError, match="LLM down"):
+            await loop.run(
+                Prompt("test"),
+                [SyntheticExample("in", id=0), SyntheticExample("in2", id=1)],
+                "task",
+            )
 
     @pytest.mark.asyncio
     async def test_resets_on_success(self):
@@ -216,12 +214,11 @@ class TestCircuitBreaker:
             circuit_breaker_threshold=3,
             error_strategy="skip",
         )
-        with patch.object(loop, "_log"):
-            state = await loop.run(
-                Prompt("test"),
-                [SyntheticExample("in", id=0), SyntheticExample("in2", id=1)],
-                "task",
-            )
+        state = await loop.run(
+            Prompt("test"),
+            [SyntheticExample("in", id=0), SyntheticExample("in2", id=1)],
+            "task",
+        )
 
         # Should NOT have tripped — 2 fails, then success reset the counter
         cb_events = [h for h in state.history if h.get("event") == "circuit_breaker"]
@@ -277,6 +274,10 @@ class TestPerCallIsolation:
         adapter._max_retries = 1
         adapter._retry_delay_base = 0
         adapter._semaphore = __import__("asyncio").Semaphore(5)
+        adapter._judge_criteria = ""
+        adapter._judge_dimensions = []
+        adapter._dimension_names = ""
+        adapter._weights = {}
 
         # Mock _judge to fail on first call, succeed on second
         call_count = 0