test: tighten 8bitlt fullgraph skip to torch < 2.5

The failure on torch 2.4.1 is dynamo failing to trace MatmulLtState
(UserDefinedObject `__bool__`), which is a different root cause than
the Params4bit `.t()` issue that needs torch 2.8+ in the 4bit test.
Torch 2.5+ has the dynamo UDO improvements needed for this to work.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: Titus-von-Koeller

tests/test_linear8bitlt.py

@@ -257,8 +257,8 @@ def test_linear8bit_load_state_dict_raises_runtime_for_tied_weight():
     torch.__version__ < (2, 10) and sys.version_info >= (3, 14), reason="Not supported in Python 3.14 until torch 2.10"
 )
 def test_linear8bitlt_torch_compile(device, threshold, bias, fullgraph, mode):
-    if fullgraph and torch.__version__ < (2, 8, 0, "dev"):
-        pytest.skip("fullgraph mode requires torch 2.8 or higher")
+    if fullgraph and torch.__version__ < (2, 5):
+        pytest.skip("fullgraph tracing of MatmulLtState requires torch >= 2.5")
 
     if device == "cuda" and platform.system() == "Windows":
         pytest.skip("Triton is not officially supported on Windows")