[Kernels] Add Flash Linear Attention Kernels (#24518)

Signed-off-by: youkaichao <youkaichao@gmail.com>

vllm/triton_utils/__init__.py

@@ -7,8 +7,10 @@ from vllm.triton_utils.importing import (HAS_TRITON, TritonLanguagePlaceholder,
 if HAS_TRITON:
     import triton
     import triton.language as tl
+    import triton.language.extra.libdevice as tldevice
 else:
     triton = TritonPlaceholder()
     tl = TritonLanguagePlaceholder()
+    tldevice = TritonLanguagePlaceholder()
 
-__all__ = ["HAS_TRITON", "triton", "tl"]
+__all__ = ["HAS_TRITON", "triton", "tl", "tldevice"]