feat: v0.2.0 sprint — ground truth eval, crossover/mutation, checkpointing, similarity guards, dataset loader, CLI commands, extended test coverage

Aggregates all v0.2.0 sprint work (GARAA-30 through GARAA-40) and fixes
2 integration tests that broke when the codebase went async (DSPyLLMAdapter
and full pipeline tests now properly await coroutines).

277 tests pass (260 unit + 17 integration).

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

src/prometheus/infrastructure/crossover_adapter.py

@@ -0,0 +1,63 @@
+"""
+Adapter: Instruction Crossover via DSPy.
+
+Implements CrossoverPort — combines two parent prompts into a child.
+"""
+from __future__ import annotations
+
+import asyncio
+
+import dspy
+
+from prometheus.domain.entities import Prompt
+from prometheus.domain.ports import CrossoverPort
+from prometheus.infrastructure.dspy_modules import InstructionCrossover
+from prometheus.infrastructure.retry import async_retry_with_backoff
+
+
+class DSPyCrossoverAdapter(CrossoverPort):
+    """Uses DSPy to combine two parent instructions into a child."""
+
+    def __init__(
+        self,
+        lm: dspy.LM,
+        max_retries: int = 3,
+        retry_delay_base: float = 1.0,
+    ) -> None:
+        self._lm = lm
+        self._crossover = InstructionCrossover()
+        self._max_retries = max_retries
+        self._retry_delay_base = retry_delay_base
+        self.call_count: int = 0
+
+    async def crossover(
+        self,
+        parent_a: Prompt,
+        parent_b: Prompt,
+        task_description: str,
+    ) -> Prompt:
+        async def _call() -> Prompt:
+            return await asyncio.to_thread(
+                self._sync_crossover, parent_a, parent_b, task_description,
+            )
+
+        return await async_retry_with_backoff(
+            _call,
+            max_retries=self._max_retries,
+            retry_delay_base=self._retry_delay_base,
+        )
+
+    def _sync_crossover(
+        self,
+        parent_a: Prompt,
+        parent_b: Prompt,
+        task_description: str,
+    ) -> Prompt:
+        with dspy.context(lm=self._lm):
+            pred = self._crossover(
+                parent_a=parent_a.text,
+                parent_b=parent_b.text,
+                task_description=task_description,
+            )
+        self.call_count += 1
+        return Prompt(text=pred.child_instruction)