refactor: deduplicate and restructure test suite

- extract preprocessing config factories into tests/data/factories.py
- keep conftest.py fixtures-only; stop importing builders from it
- promote temp_dir fixture to root conftest for cross-directory reuse
- unify duplicate BPE tokenizer builders into build_test_tokenizer
- merge grpo/dpo online e2e tests into one parametrized integration test
- extract engine mock factory and shared model batch builders
- drop local tempfile usage in favor of shared fixtures

No behavior change: 519 tests pass.

tests/data/test_preprocess_config.py

@@ -4,9 +4,9 @@ from astrai.config.preprocess_config import (
     InputConfig,
     PipelineConfig,
 )
-from tests.data.conftest import (
-    _INSTRUCTION_SECTIONS,
-    _TEXT_SECTIONS,
+from tests.data.factories import (
+    INSTRUCTION_SECTIONS,
+    TEXT_SECTIONS,
     make_dpo_chat_config,
 )
 
@@ -43,26 +43,26 @@ def test_from_dict_flat():
 
 def test_to_dict_roundtrip():
     config = PipelineConfig(
-        input=InputConfig(sections=_INSTRUCTION_SECTIONS),
+        input=InputConfig(sections=INSTRUCTION_SECTIONS),
         mask={"prompt": "mask", "response": "train"},
         mask_default="mask",
     )
     d = config.to_dict()
     config2 = PipelineConfig.from_dict(d)
-    assert config2.input.sections == _INSTRUCTION_SECTIONS
+    assert config2.input.sections == INSTRUCTION_SECTIONS
     assert config2.mask == {"prompt": "mask", "response": "train"}
 
 
 def test_to_file_from_file(temp_dir):
     config = PipelineConfig(
-        input=InputConfig(sections=_TEXT_SECTIONS),
+        input=InputConfig(sections=TEXT_SECTIONS),
         mask={"text": "train"},
         mask_default="mask",
     )
     path = os.path.join(temp_dir, "config.json")
     config.to_file(path)
     loaded = PipelineConfig.from_file(path)
-    assert loaded.input.sections == _TEXT_SECTIONS
+    assert loaded.input.sections == TEXT_SECTIONS
     assert loaded.mask == {"text": "train"}