Fix FlashAttentionDecodeSplitVx indirect dispatch input ordering

[jchen10]

Move SetIndirectDispatchTensor after all AddInput calls to ensure the
indirect buffer is the last program input.

[jchen10]

The error to fix:

WebGPU device error(2): Error while parsing WGSL: :44:23 error: no matching overload for 'operator * (u32, f16)'

9 candidate operators:
 ΓÇó 'operator * (T  Γ£ô , T  Γ£ù ) -> T' where:
      Γ£ô  'T' is 'abstract-float', 'abstract-int', 'f32', 'i32', 'u32' or 'f16'
 ΓÇó 'operator * (vecN<T>  Γ£ù , T  Γ£ô ) -> vecN<T>' where:
      Γ£ô  'T' is 'abstract-float', 'abstract-int', 'f32', 'i32', 'u32' or 'f16'
 ΓÇó 'operator * (T  Γ£ô , vecN<T>  Γ£ù ) -> vecN<T>' where:
      Γ£ô  'T' is 'abstract-float', 'abstract-int', 'f32', 'i32', 'u32' or 'f16'
 ΓÇó 'operator * (matNxM<T>  Γ£ù , T  Γ£ô ) -> matNxM<T>' where:
      Γ£ô  'T' is 'abstract-float', 'f32' or 'f16'
 ΓÇó 'operator * (T  Γ£ù , matNxM<T>  Γ£ù ) -> matNxM<T>' where:
      Γ£ù  'T' is 'abstract-float', 'f32' or 'f16'
 ΓÇó 'operator * (vecN<T>  Γ£ù , vecN<T>  Γ£ù ) -> vecN<T>' where:
      Γ£ù  'T' is 'abstract-float', 'abstract-int', 'f32', 'i32', 'u32' or 'f16'
 ΓÇó 'operator * (matCxR<T>  Γ£ù , vecC<T>  Γ£ù ) -> vecR<T>' where:
      Γ£ù  'T' is 'abstract-float', 'f32' or 'f16'
 ΓÇó 'operator * (vecR<T>  Γ£ù , matCxR<T>  Γ£ù ) -> vecC<T>' where:
      Γ£ù  'T' is 'abstract-float', 'f32' or 'f16'
 ΓÇó 'operator * (matKxR<T>  Γ£ù , matCxK<T>  Γ£ù ) -> matCxR<T>' where:
      Γ£ù  'T' is 'abstract-float', 'f32' or 'f16'

  let workgroup_idx = workgroup_id.z * num_workgroups_x * num_workgroups_y + workgroup_id.y * num_workgroups_x + workgroup_id.x;
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^


 - While calling [Device].CreateShaderModule([ShaderModuleDescriptor ""FlashAttentionDecodeSplitVx""]).

2026-04-01 14:38:15.0850117 [E:onnxruntime:, webgpu_context.cc:101 onnxruntime::webgpu::WebGpuContext::Initialize::<lambda_1>::()::<lambda_2>::operator ()] WebGPU device error(2): Error while parsing WGSL: :44:23 error: no matching overload for 'operator * (u32, f16)'

[jchen10]

@qjia7 @fs-eire @guschmue PTAL

[jchen10]

@guschmue PTAL, just a gentle reminder.

[guschmue]

/azp run Linux QNN CI Pipeline,Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 4 pipeline(s).

[guschmue]

some unrelated error with tests running on xnnpack but only on macos.
Don't see this in in other PRs.
Can you merge main?

[jchen10]

Rebased on main. Please help try again. Thanks!

[guschmue]

/azp run Linux QNN CI Pipeline,Win_TRT_Minimal_CUDA_Test_CI,Windows ARM64 QNN CI Pipeline,Windows GPU Doc Gen CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 3 pipeline(s).

[guschmue]

/azp run Windows ARM64 QNN CI Pipeline

[azure-pipelines]

No pipelines are associated with this pull request.

[eserscor]

/azp run Windows ARM64 QNN CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).