[ExecuTorch][WebGPU] SDPA: branchless aligned/tail loads in the QK/AV kernels#20510

1 commit merged into

gh/JulianCloudNTH/62/origpytorch/executorch:gh/JulianCloudNTH/62/origfrom

gh/JulianCloudNTH/63/origpytorch/executorch:gh/JulianCloudNTH/63/origCopy head branch name to clipboard