[Benchmark] template attention kernel and test

Author: Sibylau

benchmarks/run.py

@@ -280,6 +280,11 @@ class RunResult:
         "examples.jagged_sum",
         "jagged_sum_tritonbench",
     ),
+    "template_attention": (
+        "tritonbench.operators.template_attention.operator",
+        "examples.template_attention",
+        "template_attention_tritonbench",
+    ),
 }
 
 
@@ -538,6 +543,12 @@ class RunResult:
         "helion_fp8_gemm_tritonbench-speedup": "helion_speedup",
         "helion_fp8_gemm_tritonbench-accuracy": "helion_accuracy",
     },
+    "template_attention": {
+        "test_with_exp2-speedup": "triton_speedup",
+        "test_with_exp2-accuracy": "triton_accuracy",
+        "helion_template_attention_tritonbench-speedup": "helion_speedup",
+        "helion_template_attention_tritonbench-accuracy": "helion_accuracy",
+    },
 }