Apply same optimization pattern to TensorAdaptor

TensorAdaptor has identical InitializeElementSize and
GetTransformAndItsUpperDimension patterns as TensorDescriptor.
Apply the same optimization:
- Replace nested static_for lambdas with find_in_tuple_of_sequences
- Replace generate_tuple lambda with pack expansion

Results: generate_tuple lambdas 100 -> 96 (4 events, 17ms eliminated)

Author: tenpercent

include/ck/tensor_description/tensor_adaptor.hpp

@@ -45,27 +45,29 @@ struct TensorAdaptor
         return BottomDimensionHiddenIds{};
     }
 
-    __host__ __device__ static constexpr auto InitializeElementSize(const Transforms& transforms)
+    // Helper to get length of a top dimension from transforms
+    template <index_t I>
+    __host__ __device__ static constexpr auto
+    GetTopDimLengthFromTransforms(const Transforms& transforms)
     {
-        const auto lengths = generate_tuple(
-            [&](auto idim_top) {
-                constexpr auto tmp = GetTransformAndItsUpperDimension(idim_top);
-
-                constexpr index_t itran   = tmp[Number<0>{}];
-                constexpr index_t idim_up = tmp[Number<1>{}];
-                constexpr bool found      = tmp[Number<2>{}];
-
-                static_assert(found == true,
-                              "wrong! not found matching transformation and upper-dimension");
-
-                const auto length =
-                    transforms[Number<itran>{}].GetUpperLengths()[Number<idim_up>{}];
+        constexpr auto result = find_in_tuple_of_sequences<TopDimensionHiddenIds::At(Number<I>{})>(
+            UpperDimensionHiddenIdss{});
+        static_assert(result.found, "wrong! not found matching transformation and upper-dimension");
+        return transforms[Number<result.itran>{}].GetUpperLengths()[Number<result.idim_up>{}];
+    }
 
-                return length;
-            },
-            Number<ndim_top_>{});
+    // Compute element size using pack expansion instead of generate_tuple with lambda
+    template <index_t... Is>
+    __host__ __device__ static constexpr auto ComputeElementSizeImpl(const Transforms& transforms,
+                                                                     Sequence<Is...>)
+    {
+        return (GetTopDimLengthFromTransforms<Is>(transforms) * ...);
+    }
 
-        return container_product(lengths);
+    __host__ __device__ static constexpr auto InitializeElementSize(const Transforms& transforms)
+    {
+        return ComputeElementSizeImpl(transforms,
+                                      typename arithmetic_sequence_gen<0, ndim_top_, 1>::type{});
     }
 
     template <index_t IDim>
@@ -75,24 +77,10 @@ struct TensorAdaptor
 
         constexpr index_t idim_hidden = TopDimensionHiddenIds::At(idim_top);
 
-        index_t itran_found   = 0;
-        index_t idim_up_found = 0;
-        bool found            = false;
-
-        static_for<0, ntransform_, 1>{}([&](auto itran) {
-            constexpr auto up_dim_ids = UpperDimensionHiddenIdss{}[itran];
-
-            static_for<0, up_dim_ids.Size(), 1>{}([&](auto idim_up) {
-                if constexpr(up_dim_ids[idim_up] == idim_hidden)
-                {
-                    itran_found   = itran;
-                    idim_up_found = idim_up;
-                    found         = true;
-                }
-            });
-        });
+        // Use compile-time search helper instead of nested static_for with lambdas
+        constexpr auto result = find_in_tuple_of_sequences<idim_hidden>(UpperDimensionHiddenIdss{});
 
-        return make_tuple(itran_found, idim_up_found, found);
+        return make_tuple(result.itran, result.idim_up, result.found);
     }
 
     __host__ __device__ static constexpr index_t GetNumOfBottomDimension()