feat(llm): añade Protocol LLMProvider con Message y CompletionResult

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

tests/unit/test_llm_base.py

@@ -0,0 +1,19 @@
+"""Test de los modelos base del proveedor LLM."""
+
+from agentforge_core.llm.base import CompletionResult, Message
+
+
+def test_message_serializable() -> None:
+    m = Message(role="user", content="hola")
+    assert m.model_dump()["role"] == "user"
+
+
+def test_completion_result_campos_basicos() -> None:
+    r = CompletionResult(
+        content='{"a": 1}',
+        model="gpt-4o-mock",
+        tokens_in=10,
+        tokens_out=20,
+        latency_ms=42,
+    )
+    assert r.tokens_in == 10