test: deduplicate suites and prune low-value cases

- extract shared helpers for dataset writers, scheduler construction, thread interleaving, hf roundtrips, and moe configs
- remove about 20 cases whose only assertions were format checks, restated declarations, fake-taxonomy duplicates, or test-local scaffolding
- strengthen weak cases into exact reference comparisons, positional mask checks, and deterministic outcomes
- replace two schedule factory smoke tests with cosine/sgdr formula assertions
- delete root-level CLI tests whose merge-priority facts are covered by tests/config/test_cli.py
- suite shrinks from 857 to 826 items; ruff format, import order, and pytest all green

tests/module/test_forward_configs.py

@@ -369,7 +369,9 @@ def test_moe_component_forward_returns_ffn_output():
     output = moe(torch.randn(2, 8, 8))
 
     assert output["hidden_states"].shape == (2, 8, 8)
-    assert output["aux_loss"] is not None
+    assert output["aux_loss"].ndim == 0
+    assert output["aux_loss"].requires_grad
+    assert torch.isfinite(output["aux_loss"])
 
 
 @pytest.mark.parametrize("decoder_sparse_step", [0, -1])

tests/module/test_lora.py

@@ -332,4 +332,6 @@ def test_collect_lora_info():
     info = _collect_lora_info(model)
     assert "q_proj" in info
     assert "o_proj" in info
-    assert "q_proj" in info  # each layer has one
+    # every decoder layer contributes one of each attention projection
+    assert len(info["q_proj"]) == len(info["o_proj"]) == 2
+    assert all(name.endswith("attention.q_proj") for name in info["q_proj"])