feat(config): add enable_thinking flag (default False) + fix HMMT bench gold answers

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

tests/test_config.py

@@ -74,3 +74,13 @@ def test_final_tokens_zero_rejected():
 def test_final_tokens_negative_rejected():
     with pytest.raises(ValueError, match="final_tokens must be positive"):
         RSAConfig(final_tokens=-1)
+
+
+def test_enable_thinking_default_false():
+    cfg = RSAConfig()
+    assert cfg.enable_thinking is False
+
+
+def test_enable_thinking_can_be_enabled():
+    cfg = RSAConfig(enable_thinking=True)
+    assert cfg.enable_thinking is True